Dataset Preview
Duplicate
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 2 new columns ({'n', 'accuracy'}) and 32 missing columns ({'timestamp', 'total_prompt_tokens', 'mean_cost', 'token_source_counts', 'mean_search_latency', 'mean_latency', 'mean_prompt_tokens', 'rag_generator_model', 'mean_rerank_latency', 'mean_answer_relevancy', 'n_cost_known', 'total_tokens', 'mean_answer_correctness', 'total_cost', 'corpus', 'rag_generator_name', 'mean_context_recall', 'n_questions', 'total_completion_tokens', 'by_format', 'weighted_score_pct', 'model', 'mean_total_tokens', 'mean_judge_prompt_tokens', 'total_judge_prompt_tokens', 'mean_completion_tokens', 'mean_faithfulness', 'total_judge_completion_tokens', 'top_k', 'mean_accuracy', 'mean_judge_completion_tokens', 'mean_gen_latency'}).

This happened while the json dataset builder was generating data using

hf://datasets/LiamDuero/telco-analysis/mc_results.json (at revision d0c13bfd8607d67a66d7371492bceb13b5ed9944), ['hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/all_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/mc_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/open_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/tf_results.json'], ['hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/all_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/mc_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/open_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/tf_results.json']

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1848, in _prepare_split_single
                  writer.write_table(table)
                  ~~~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
                  self._write_table(pa_table, writer_batch_size=writer_batch_size)
                  ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
                  pa_table = table_cast(pa_table, self._schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              model_id: string
              model_display_name: string
              phase: string
              reranker_label: string
              reranker_key: string
              run_name: string
              questions_file: string
              n: int64
              mean_similarity: double
              mean_distinct_sources_retrieved: double
              canonical_hit_rate: double
              n_canonical_source_non_discriminating: int64
              exact_chunk_hit_rate: double
              mean_coverage_ratio: double
              accuracy: double
              mean_accuracy_attempted: double
              abstention_rate: double
              wrong_rate: double
              to
              {'model_id': Value('string'), 'model_display_name': Value('string'), 'phase': Value('string'), 'reranker_label': Value('string'), 'reranker_key': Value('string'), 'run_name': Value('string'), 'corpus': Value('string'), 'model': Value('string'), 'top_k': Value('int64'), 'timestamp': Value('string'), 'rag_generator_model': Value('string'), 'rag_generator_name': Value('string'), 'questions_file': Value('string'), 'n_questions': Value('int64'), 'mean_accuracy': Value('float64'), 'mean_accuracy_attempted': Value('float64'), 'abstention_rate': Value('float64'), 'wrong_rate': Value('float64'), 'mean_faithfulness': Value('float64'), 'mean_answer_relevancy': Value('float64'), 'mean_similarity': Value('float64'), 'mean_distinct_sources_retrieved': Value('float64'), 'canonical_hit_rate': Value('float64'), 'n_canonical_source_non_discriminating': Value('int64'), 'exact_chunk_hit_rate': Value('float64'), 'mean_coverage_ratio': Value('float64'), 'mean_context_recall': Value('float64'), 'mean_answer_correctness': Value('float64'), 'mean_latency': Value('float64'), 'mean_search_latency': Value('float64'), 'mean_rerank_latency': Value('float64'), 'mean_gen_latency': Value('float64'), 'total_prompt_tokens': Value('int64'), 'total_completion_tokens': Value('int64'), 'total_tokens': Value('int64'), 'mean_prompt_tokens': Value('float64'), 'mean_completion_tokens': Value('float64'), 'mean_total_tokens': Value('float64'), 'total_judge_prompt_tokens': Value('int64'), 'total_judge_completion_tokens':
              ...
              'mean_cost': Value('float64'), 'n_cost_known': Value('int64'), 'token_source_counts': {'api': Value('int64')}, 'weighted_score_pct': Value('float64'), 'by_format': {'mc': {'n': Value('int64'), 'mean_similarity': Value('float64'), 'mean_distinct_sources_retrieved': Value('float64'), 'canonical_hit_rate': Value('float64'), 'n_canonical_source_non_discriminating': Value('int64'), 'exact_chunk_hit_rate': Value('float64'), 'mean_coverage_ratio': Value('float64'), 'accuracy': Value('float64'), 'mean_accuracy_attempted': Value('float64'), 'abstention_rate': Value('float64'), 'wrong_rate': Value('float64')}, 'open': {'n': Value('int64'), 'mean_similarity': Value('float64'), 'mean_distinct_sources_retrieved': Value('float64'), 'canonical_hit_rate': Value('float64'), 'n_canonical_source_non_discriminating': Value('int64'), 'exact_chunk_hit_rate': Value('float64'), 'mean_coverage_ratio': Value('float64'), 'faithfulness': Value('float64'), 'answer_relevancy': Value('float64'), 'context_recall': Value('float64'), 'answer_correctness': Value('float64')}, 'tf': {'n': Value('int64'), 'mean_similarity': Value('float64'), 'mean_distinct_sources_retrieved': Value('float64'), 'canonical_hit_rate': Value('float64'), 'n_canonical_source_non_discriminating': Value('int64'), 'exact_chunk_hit_rate': Value('float64'), 'mean_coverage_ratio': Value('float64'), 'accuracy': Value('float64'), 'mean_accuracy_attempted': Value('float64'), 'abstention_rate': Value('float64'), 'wrong_rate': Value('float64')}}}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1850, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
                  ...<4 lines>...
                  )
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 2 new columns ({'n', 'accuracy'}) and 32 missing columns ({'timestamp', 'total_prompt_tokens', 'mean_cost', 'token_source_counts', 'mean_search_latency', 'mean_latency', 'mean_prompt_tokens', 'rag_generator_model', 'mean_rerank_latency', 'mean_answer_relevancy', 'n_cost_known', 'total_tokens', 'mean_answer_correctness', 'total_cost', 'corpus', 'rag_generator_name', 'mean_context_recall', 'n_questions', 'total_completion_tokens', 'by_format', 'weighted_score_pct', 'model', 'mean_total_tokens', 'mean_judge_prompt_tokens', 'total_judge_prompt_tokens', 'mean_completion_tokens', 'mean_faithfulness', 'total_judge_completion_tokens', 'top_k', 'mean_accuracy', 'mean_judge_completion_tokens', 'mean_gen_latency'}).
              
              This happened while the json dataset builder was generating data using
              
              hf://datasets/LiamDuero/telco-analysis/mc_results.json (at revision d0c13bfd8607d67a66d7371492bceb13b5ed9944), ['hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/all_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/mc_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/open_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/tf_results.json'], ['hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/all_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/mc_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/open_results.json', 'hf://datasets/LiamDuero/telco-analysis@d0c13bfd8607d67a66d7371492bceb13b5ed9944/tf_results.json']
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

model_id
string
model_display_name
string
phase
string
reranker_label
string
reranker_key
string
run_name
string
corpus
string
model
string
top_k
int64
timestamp
string
rag_generator_model
string
rag_generator_name
string
questions_file
string
n_questions
int64
mean_accuracy
float64
mean_accuracy_attempted
float64
abstention_rate
float64
wrong_rate
float64
mean_faithfulness
float64
mean_answer_relevancy
float64
mean_similarity
float64
mean_distinct_sources_retrieved
float64
canonical_hit_rate
float64
n_canonical_source_non_discriminating
int64
exact_chunk_hit_rate
float64
mean_coverage_ratio
float64
mean_context_recall
float64
mean_answer_correctness
float64
mean_latency
float64
mean_search_latency
float64
mean_rerank_latency
float64
mean_gen_latency
float64
total_prompt_tokens
int64
total_completion_tokens
int64
total_tokens
int64
mean_prompt_tokens
float64
mean_completion_tokens
float64
mean_total_tokens
float64
total_judge_prompt_tokens
int64
total_judge_completion_tokens
int64
mean_judge_prompt_tokens
float64
mean_judge_completion_tokens
float64
total_cost
float64
mean_cost
float64
n_cost_known
int64
token_source_counts
dict
weighted_score_pct
float64
by_format
dict
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
10
20260827_204848
google/gemma-4-31b-it
local
custom_list
105
0.757895
0.986301
0.231579
0.010526
0.755
0.577056
0.745114
1
0.87619
0
0.571429
1
0.676667
0.707631
7.284846
2.066095
0.332203
4.886526
153,681
25,143
178,824
1,463.628571
239.457143
1,703.085714
46,534
6,321
443.180952
60.2
0
0
105
{ "api": 105 }
66.896552
{ "mc": { "n": 55, "mean_similarity": 0.7298490909, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8909090909, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.5454545455, "mean_coverage_ratio": 1, "accuracy": 0.7090909091, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
10
20260827_205132
google/gemma-4-31b-it
local
custom_list
105
0.631579
0.967742
0.347368
0.021053
0.546548
0.428701
0.750733
1
0.752381
0
0.228571
1
0.512143
0.64557
6.770276
0.107498
0.028193
6.634563
405,825
26,894
432,719
3,865
256.133333
4,121.133333
54,377
4,816
517.87619
45.866667
0
0
105
{ "api": 105 }
51.724138
{ "mc": { "n": 55, "mean_similarity": 0.7408636364, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.5636363636, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
10
20260827_205346
google/gemma-4-31b-it
local
custom_list
105
0.631579
0.967742
0.347368
0.021053
0.670668
0.53341
0.759572
1
0.838095
0
0.066667
1
0.545
0.608506
4.574135
0.106684
0.016034
4.451397
151,416
23,550
174,966
1,442.057143
224.285714
1,666.342857
40,041
5,486
381.342857
52.247619
0
0
105
{ "api": 105 }
51.724138
{ "mc": { "n": 55, "mean_similarity": 0.7469690909, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8363636364, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0909090909, "mean_coverage_ratio": 1, "accuracy": 0.6181818182, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
10
20260827_205525
google/gemma-4-31b-it
local
custom_list
105
0.663158
0.969231
0.315789
0.021053
0.728611
0.530448
0.696667
1
0.790476
0
0.266667
1
0.55
0.576442
5.944751
0.071437
0.016672
5.856621
242,040
25,963
268,003
2,305.142857
247.266667
2,552.409524
57,011
5,074
542.961905
48.32381
0
0
105
{ "api": 105 }
53.793103
{ "mc": { "n": 55, "mean_similarity": 0.6991090909, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.5818181818, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
10
20260827_205733
google/gemma-4-31b-it
local
custom_list
105
0.684211
1
0.315789
0
0.577778
0.441754
0.77625
1
0.790476
0
0.504762
1
0.583333
0.604842
6.587285
1.255191
0.409207
4.922865
154,713
24,273
178,986
1,473.457143
231.171429
1,704.628571
40,191
5,010
382.771429
47.714286
0
0
105
{ "api": 105 }
62.068966
{ "mc": { "n": 55, "mean_similarity": 0.7615618182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4, "mean_coverage_ratio": 1, "accuracy": 0.6, "mean_accuracy_attempted": 1, "abste...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
10
20260827_204627
google/gemma-4-31b-it
local
custom_list
105
0.694737
1
0.305263
0
0.566667
0.498385
0.848995
1
0.809524
0
0.514286
1
0.608333
0.651607
6.46782
1.157041
0.387448
4.923308
153,303
24,901
178,204
1,460.028571
237.152381
1,697.180952
38,105
4,886
362.904762
46.533333
0
0
105
{ "api": 105 }
62.758621
{ "mc": { "n": 55, "mean_similarity": 0.8427509091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4363636364, "mean_coverage_ratio": 1, "accuracy": 0.6545454545, "mean_accuracy_attempted": 1, ...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
10
20260827_210205
google/gemma-4-31b-it
local
custom_list
105
0.789474
0.949367
0.168421
0.042105
0.887427
0.602587
0.745806
1
0.838095
0
0.590476
1
0.657143
0.686898
7.985103
2.012687
0.937207
5.035188
160,950
25,077
186,027
1,532.857143
238.828571
1,771.685714
46,418
6,499
442.07619
61.895238
0
0
105
{ "api": 105 }
72.413793
{ "mc": { "n": 55, "mean_similarity": 0.73362, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8545454545, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.5454545455, "mean_coverage_ratio": 1, "accuracy": 0.7818181818, "mean_accuracy_attempted": ...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
10
20260827_210457
google/gemma-4-31b-it
local
custom_list
105
0.484211
0.901961
0.463158
0.052632
0.570833
0.443833
0.73727
1
0.714286
0
0.171429
1
0.405714
0.549609
8.353087
0.100173
0.076669
8.176224
609,870
27,354
637,224
5,808.285714
260.514286
6,068.8
45,000
4,773
428.571429
45.457143
0
0
105
{ "api": 105 }
38.62069
{ "mc": { "n": 55, "mean_similarity": 0.7374418182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2363636364, "mean_coverage_ratio": 1, "accuracy": 0.5454545455, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
10
20260827_210712
google/gemma-4-31b-it
local
custom_list
105
0.652632
0.96875
0.326316
0.021053
0.76412
0.531918
0.761608
1
0.809524
0
0.028571
1
0.717857
0.69662
5.024463
0.101414
0.074125
4.8489
158,994
25,054
184,048
1,514.228571
238.609524
1,752.838095
42,270
6,224
402.571429
59.27619
0
0
105
{ "api": 105 }
63.448276
{ "mc": { "n": 55, "mean_similarity": 0.7437854545, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7454545455, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0363636364, "mean_coverage_ratio": 1, "accuracy": 0.6, "mean_accuracy_attempted": 0.97...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
10
20260827_210853
google/gemma-4-31b-it
local
custom_list
105
0.442105
0.893617
0.505263
0.052632
0.546296
0.422603
0.685535
1
0.742857
0
0.180952
1
0.593095
0.617864
6.120315
0.073409
0.074789
5.972095
244,083
26,379
270,462
2,324.6
251.228571
2,575.828571
52,631
4,724
501.247619
44.990476
0
0
105
{ "api": 105 }
42.758621
{ "mc": { "n": 55, "mean_similarity": 0.6885454545, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7090909091, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.4909090909, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
10
20260827_211100
google/gemma-4-31b-it
local
custom_list
105
0.715789
0.985507
0.273684
0.010526
0.851501
0.622586
0.774981
1
0.8
0
0.561905
1
0.65
0.676632
7.171008
1.256393
0.808419
5.106175
158,517
25,017
183,534
1,509.685714
238.257143
1,747.942857
44,546
6,131
424.247619
58.390476
0
0
105
{ "api": 105 }
64.137931
{ "mc": { "n": 55, "mean_similarity": 0.76274, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4909090909, "mean_coverage_ratio": 1, "accuracy": 0.6909090909, "mean_accuracy_attempted": ...
google/gemma-4-31b-it
Gemma 4 31B (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
10
20260827_205945
google/gemma-4-31b-it
local
custom_list
105
0.684211
0.984848
0.305263
0.010526
0.824041
0.743309
0.845741
1
0.790476
0
0.542857
1
0.775
0.76836
7.293818
1.056007
1.182867
5.054923
160,089
25,285
185,374
1,524.657143
240.809524
1,765.466667
46,909
6,517
446.752381
62.066667
0
0
105
{ "api": 105 }
65.517241
{ "mc": { "n": 55, "mean_similarity": 0.8396581818, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7454545455, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4727272727, "mean_coverage_ratio": 1, "accuracy": 0.6363636364, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
no_reranker
null
null
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
1
20260827_223954
google/gemma-4-31b-it
local
custom_list
105
0.747368
0.972603
0.231579
0.021053
0.781871
0.54769
0.760254
1
0.828571
0
0.571429
1
0.6
0.624425
6.901312
1.943326
0
4.957966
149,649
24,867
174,516
1,425.228571
236.828571
1,662.057143
42,179
5,512
401.704762
52.495238
0
0
105
{ "api": 105 }
62.758621
{ "mc": { "n": 55, "mean_similarity": 0.7510145455, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4909090909, "mean_coverage_ratio": 1, "accuracy": 0.6545454545, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
no_reranker
null
null
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
1
20260827_224256
google/gemma-4-31b-it
local
custom_list
105
0.736842
0.921053
0.2
0.063158
0.831972
0.632688
0.786716
1
0.857143
0
0.361905
1
0.622381
0.701495
8.67485
0.093499
0
8.581331
610,458
28,386
638,844
5,813.885714
270.342857
6,084.228571
64,562
6,182
614.87619
58.87619
0
0
105
{ "api": 105 }
68.965517
{ "mc": { "n": 55, "mean_similarity": 0.7705709091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8363636364, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.3454545455, "mean_coverage_ratio": 1, "accuracy": 0.6909090909, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
no_reranker
null
null
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
1
20260827_224511
google/gemma-4-31b-it
local
custom_list
105
0.621053
0.967213
0.357895
0.021053
0.722751
0.531402
0.777741
1
0.847619
0
0.047619
1
0.545714
0.630537
4.676933
0.113013
0
4.5639
148,944
23,610
172,554
1,418.514286
224.857143
1,643.371429
38,741
4,896
368.961905
46.628571
0
0
105
{ "api": 105 }
57.931034
{ "mc": { "n": 55, "mean_similarity": 0.76122, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8181818182, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0727272727, "mean_coverage_ratio": 1, "accuracy": 0.5090909091, "mean_accuracy_attempted": ...
google/gemma-4-31b-it
Gemma 4 31B (local)
no_reranker
null
null
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
1
20260827_224655
google/gemma-4-31b-it
local
custom_list
105
0.610526
0.966667
0.368421
0.021053
0.516944
0.411953
0.714388
1
0.742857
0
0.247619
1
0.561667
0.615785
6.073447
0.067771
0
6.005656
239,310
26,619
265,929
2,279.142857
253.514286
2,532.657143
50,591
4,885
481.819048
46.52381
0
0
105
{ "api": 105 }
57.241379
{ "mc": { "n": 55, "mean_similarity": 0.71458, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7272727273, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2545454545, "mean_coverage_ratio": 1, "accuracy": 0.6, "mean_accuracy_attempted": 0.9705882...
google/gemma-4-31b-it
Gemma 4 31B (local)
no_reranker
null
null
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
1
20260827_224857
google/gemma-4-31b-it
local
custom_list
105
0.578947
0.948276
0.389474
0.031579
0.658333
0.494843
0.789144
1
0.72381
0
0.457143
1
0.566667
0.61056
6.292078
1.266316
0
5.025742
146,043
24,646
170,689
1,390.885714
234.72381
1,625.609524
39,934
5,935
380.32381
56.52381
0
0
105
{ "api": 105 }
51.724138
{ "mc": { "n": 55, "mean_similarity": 0.7770309091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.6545454545, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.3636363636, "mean_coverage_ratio": 1, "accuracy": 0.4909090909, "mean_accuracy_attempt...
google/gemma-4-31b-it
Gemma 4 31B (local)
no_reranker
null
null
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
1
20260827_223746
google/gemma-4-31b-it
local
custom_list
105
0.589474
0.949153
0.378947
0.031579
0.561738
0.500938
0.854764
1
0.771429
0
0.47619
1
0.608333
0.635459
6.194986
1.08371
0
5.111256
161,514
25,310
186,824
1,538.228571
241.047619
1,779.27619
38,362
5,316
365.352381
50.628571
0
0
105
{ "api": 105 }
55.862069
{ "mc": { "n": 55, "mean_similarity": 0.8493709091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4, "mean_coverage_ratio": 1, "accuracy": 0.5454545455, "mean_accuracy_attempted": 0.93...
google/gemma-4-31b-it
Gemma 4 31B (local)
baseline
null
null
3gpp_baseline_google-gemma-4-31b-it_no_retrieval_final_3gpp_105
3gpp
baseline-google-gemma-4-31b-it
0
20260828_093342
final_3gpp_105.json
105
0.463158
0.463158
0
0.536842
0
0.707718
0
0
-1
0
-1
-1
0
0.605929
3.724855
0
0
3.724839
22,025
20,649
42,674
209.761905
196.657143
406.419048
14,467
1,101
137.780952
10.485714
null
null
0
{ "api": 105 }
40.689655
{ "mc": { "n": 55, "mean_similarity": 0, "mean_distinct_sources_retrieved": 0, "canonical_hit_rate": -1, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": -1, "mean_coverage_ratio": -1, "accuracy": 0.5272727273, "mean_accuracy_attempted": 0.5272727273, "absten...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
10
20260828_072048
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.778947
0.936709
0.168421
0.052632
0.768853
0.652655
0.745114
1
0.87619
0
0.571429
1
0.665833
0.685123
10.565077
1.838705
0.279263
8.447087
152,250
41,930
194,180
1,450
399.333333
1,849.333333
78,353
6,572
746.219048
62.590476
0
0
105
{ "api": 105 }
68.275862
{ "mc": { "n": 55, "mean_similarity": 0.7298490909, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8909090909, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.5454545455, "mean_coverage_ratio": 1, "accuracy": 0.7636363636, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
10
20260828_072458
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.715789
0.918919
0.221053
0.063158
0.610304
0.536133
0.750724
1
0.752381
0
0.228571
1
0.512143
0.658943
15.073881
0.085656
0.023135
14.965068
408,468
44,811
453,279
3,890.171429
426.771429
4,316.942857
86,854
5,543
827.180952
52.790476
0
0
105
{ "api": 105 }
60.689655
{ "mc": { "n": 55, "mean_similarity": 0.7408418182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.7272727273, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
10
20260828_072735
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.757895
0.923077
0.178947
0.063158
0.641111
0.498525
0.759573
1
0.838095
0
0.066667
1
0.545
0.6595
7.995698
0.093926
0.016337
7.885413
149,196
40,462
189,658
1,420.914286
385.352381
1,806.266667
62,800
4,849
598.095238
46.180952
0
0
105
{ "api": 105 }
66.896552
{ "mc": { "n": 55, "mean_similarity": 0.7469654545, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8363636364, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0909090909, "mean_coverage_ratio": 1, "accuracy": 0.7636363636, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
10
20260828_073001
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.736842
0.958904
0.231579
0.031579
0.612037
0.496663
0.696654
1
0.790476
0
0.266667
1
0.55
0.622462
9.396753
0.066273
0.016471
9.313985
240,105
42,925
283,030
2,286.714286
408.809524
2,695.52381
80,113
5,064
762.980952
48.228571
0
0
105
{ "api": 105 }
65.517241
{ "mc": { "n": 55, "mean_similarity": 0.6990963636, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.7090909091, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
10
20260828_073242
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.705263
0.957143
0.263158
0.031579
0.433532
0.359276
0.77625
1
0.790476
0
0.504762
1
0.583333
0.604056
9.697122
1.134635
0.352352
8.210112
153,459
40,254
193,713
1,461.514286
383.371429
1,844.885714
53,716
3,825
511.580952
36.428571
0
0
105
{ "api": 105 }
60
{ "mc": { "n": 55, "mean_similarity": 0.7615618182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4, "mean_coverage_ratio": 1, "accuracy": 0.6727272727, "mean_accuracy_attempted": 0.92...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
10
20260828_071735
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.715789
0.931507
0.231579
0.052632
0.656905
0.612839
0.848995
1
0.809524
0
0.514286
1
0.608333
0.675668
9.824407
1.067066
0.342115
8.415202
151,638
41,137
192,775
1,444.171429
391.780952
1,835.952381
63,022
5,214
600.209524
49.657143
0
0
105
{ "api": 105 }
64.137931
{ "mc": { "n": 55, "mean_similarity": 0.8427509091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4363636364, "mean_coverage_ratio": 1, "accuracy": 0.6909090909, "mean_accuracy_attempted": 0.88...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
10
20260828_073846
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.8
0.926829
0.136842
0.063158
0.732621
0.610606
0.745797
1
0.838095
0
0.590476
1
0.657143
0.67835
10.894233
1.748882
0.715446
8.429883
159,225
41,609
200,834
1,516.428571
396.27619
1,912.704762
75,939
5,934
723.228571
56.514286
0
0
105
{ "api": 105 }
69.655172
{ "mc": { "n": 55, "mean_similarity": 0.7336163636, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8545454545, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.5454545455, "mean_coverage_ratio": 1, "accuracy": 0.7818181818, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
10
20260828_074332
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.578947
0.873016
0.336842
0.084211
0.475952
0.447162
0.737269
1
0.714286
0
0.171429
1
0.405714
0.564725
18.812156
0.091307
0.070559
18.650267
595,914
41,407
637,321
5,675.371429
394.352381
6,069.72381
64,706
4,849
616.247619
46.180952
0
0
105
{ "api": 105 }
48.275862
{ "mc": { "n": 55, "mean_similarity": 0.7374436364, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2363636364, "mean_coverage_ratio": 1, "accuracy": 0.6727272727, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
10
20260828_074607
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.757895
0.9
0.157895
0.084211
0.638333
0.533424
0.761595
1
0.809524
0
0.028571
1
0.717857
0.713832
8.173429
0.097557
0.070982
8.004867
156,273
40,284
196,557
1,488.314286
383.657143
1,871.971429
63,892
5,390
608.495238
51.333333
0
0
105
{ "api": 105 }
77.241379
{ "mc": { "n": 55, "mean_similarity": 0.7437709091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7454545455, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0363636364, "mean_coverage_ratio": 1, "accuracy": 0.7272727273, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
10
20260828_074854
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.568421
0.915254
0.378947
0.052632
0.545803
0.487169
0.685535
1
0.742857
0
0.180952
1
0.588095
0.625584
10.613854
0.069022
0.069405
10.475405
241,791
49,050
290,841
2,302.771429
467.142857
2,769.914286
90,178
5,722
858.838095
54.495238
0
0
105
{ "api": 105 }
51.034483
{ "mc": { "n": 55, "mean_similarity": 0.6885454545, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7090909091, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.6181818182, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
10
20260828_075159
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.778947
0.986667
0.210526
0.010526
0.741488
0.605419
0.774981
1
0.8
0
0.561905
1
0.65
0.705226
10.824466
1.277414
0.969377
8.577654
156,879
42,011
198,890
1,494.085714
400.104762
1,894.190476
73,996
6,471
704.72381
61.628571
0
0
105
{ "api": 105 }
71.724138
{ "mc": { "n": 55, "mean_similarity": 0.76274, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4909090909, "mean_coverage_ratio": 1, "accuracy": 0.7636363636, "mean_accuracy_attempted": ...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
10
20260828_073547
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.736842
0.933333
0.210526
0.052632
0.830263
0.783256
0.845741
1
0.790476
0
0.542857
1
0.775
0.767596
10.607199
1.073054
0.827712
8.706411
158,109
43,002
201,111
1,505.8
409.542857
1,915.342857
81,067
7,092
772.066667
67.542857
0
0
105
{ "api": 105 }
75.862069
{ "mc": { "n": 55, "mean_similarity": 0.8396581818, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7454545455, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4727272727, "mean_coverage_ratio": 1, "accuracy": 0.6909090909, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
no_reranker
null
null
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
1
20260828_081359
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.768421
0.9125
0.157895
0.073684
0.702317
0.577187
0.760262
1
0.828571
0
0.571429
1
0.6
0.658065
10.17624
1.757171
0
8.419045
148,170
41,862
190,032
1,411.142857
398.685714
1,809.828571
72,867
6,235
693.971429
59.380952
0
0
105
{ "api": 105 }
67.586207
{ "mc": { "n": 55, "mean_similarity": 0.75102, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4909090909, "mean_coverage_ratio": 1, "accuracy": 0.6909090909, "mean_accuracy_attempted": ...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
no_reranker
null
null
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
1
20260828_081856
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.789474
0.925926
0.147368
0.063158
0.636296
0.614484
0.786725
1
0.857143
0
0.361905
1
0.622381
0.687081
19.564887
0.087684
0
19.477181
604,695
41,194
645,889
5,759
392.32381
6,151.32381
85,638
6,049
815.6
57.609524
0
0
105
{ "api": 105 }
72.413793
{ "mc": { "n": 55, "mean_similarity": 0.7705836364, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8363636364, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.3454545455, "mean_coverage_ratio": 1, "accuracy": 0.7454545455, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
no_reranker
null
null
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
1
20260828_082129
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.768421
0.890244
0.136842
0.094737
0.580238
0.464837
0.777742
1
0.847619
0
0.047619
1
0.548571
0.625475
8.218802
0.102155
0
8.116627
146,403
40,699
187,102
1,394.314286
387.609524
1,781.92381
58,748
4,454
559.504762
42.419048
0
0
105
{ "api": 105 }
64.137931
{ "mc": { "n": 55, "mean_similarity": 0.7612163636, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8181818182, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0727272727, "mean_coverage_ratio": 1, "accuracy": 0.6909090909, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
no_reranker
null
null
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
1
20260828_082355
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.663158
0.913043
0.273684
0.063158
0.563508
0.437636
0.714392
1
0.742857
0
0.247619
1
0.554048
0.648587
10.150544
0.063156
0
10.087366
237,156
47,321
284,477
2,258.628571
450.67619
2,709.304762
82,116
5,274
782.057143
50.228571
0
0
105
{ "api": 105 }
57.241379
{ "mc": { "n": 55, "mean_similarity": 0.7145909091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7272727273, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2545454545, "mean_coverage_ratio": 1, "accuracy": 0.6909090909, "mean_accuracy_attempt...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
no_reranker
null
null
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
1
20260828_082630
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.663158
0.969231
0.315789
0.021053
0.516283
0.432626
0.789144
1
0.72381
0
0.457143
1
0.566667
0.609916
9.460259
1.330367
0
8.129872
144,864
40,341
185,205
1,379.657143
384.2
1,763.857143
56,596
5,118
539.009524
48.742857
0
0
105
{ "api": 105 }
60.689655
{ "mc": { "n": 55, "mean_similarity": 0.7770309091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.6545454545, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.3636363636, "mean_coverage_ratio": 1, "accuracy": 0.6, "mean_accuracy_attempted": 0.94...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
no_reranker
null
null
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
1
20260828_081115
KU-DFI/TelecomGPT-R1
local
custom_list
105
0.642105
0.953125
0.326316
0.031579
0.446078
0.437513
0.854764
1
0.771429
0
0.47619
1
0.608333
0.657297
9.73192
1.061632
0
8.670267
159,315
40,929
200,244
1,517.285714
389.8
1,907.085714
53,469
4,297
509.228571
40.92381
0
0
105
{ "api": 105 }
59.310345
{ "mc": { "n": 55, "mean_similarity": 0.8493709091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4, "mean_coverage_ratio": 1, "accuracy": 0.6363636364, "mean_accuracy_attempted": 0.94...
KU-DFI/TelecomGPT-R1
TelecomGPT-R1 (local)
baseline
null
null
3gpp_baseline_KU-DFI-TelecomGPT-R1_no_retrieval_final_3gpp_105
3gpp
baseline-KU-DFI-TelecomGPT-R1
0
20260828_085206
final_3gpp_105.json
105
0.189474
0.195652
0.031579
0.778947
0
0.728947
0
0
-1
0
-1
-1
0
0.591893
11.53733
0
0
11.537309
21,692
68,465
90,157
206.590476
652.047619
858.638095
14,200
1,180
135.238095
11.238095
null
null
0
{ "api": 105 }
22.758621
{ "mc": { "n": 55, "mean_similarity": 0, "mean_distinct_sources_retrieved": 0, "canonical_hit_rate": -1, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": -1, "mean_coverage_ratio": -1, "accuracy": 0.1272727273, "mean_accuracy_attempted": 0.1272727273, "absten...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
10
20260827_230413
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.705263
0.905405
0.221053
0.073684
0.583466
0.397909
0.745105
1
0.87619
0
0.571429
1
0.663333
0.559879
30.903671
2.062545
0.373445
28.467588
163,195
117,515
280,710
1,554.238095
1,119.190476
2,673.428571
41,800
4,798
398.095238
45.695238
0.47015
0.004478
105
{ "api": 105 }
56.551724
{ "mc": { "n": 55, "mean_similarity": 0.7298418182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8909090909, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.5454545455, "mean_coverage_ratio": 1, "accuracy": 0.6727272727, "mean_accuracy_attempt...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
10
20260827_231042
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.589474
0.823529
0.284211
0.126316
0.415251
0.331579
0.750717
1
0.752381
0
0.228571
1
0.513095
0.532028
24.48887
0.094963
0.023066
24.370764
409,390
123,619
533,009
3,898.952381
1,177.32381
5,076.27619
51,063
4,211
486.314286
40.104762
0.646092
0.006153
105
{ "api": 105 }
45.517241
{ "mc": { "n": 55, "mean_similarity": 0.7408418182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.5272727273, "mean_accuracy_attempt...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
10
20260827_231520
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.589474
0.823529
0.284211
0.126316
0.574444
0.377815
0.759568
1
0.838095
0
0.066667
1
0.545
0.549675
20.980266
0.094151
0.014203
20.871822
154,430
114,873
269,303
1,470.761905
1,094.028571
2,564.790476
36,264
3,961
345.371429
37.72381
0.410207
0.003907
105
{ "api": 105 }
48.965517
{ "mc": { "n": 55, "mean_similarity": 0.74696, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8363636364, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0909090909, "mean_coverage_ratio": 1, "accuracy": 0.5636363636, "mean_accuracy_attempted": ...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
10
20260827_232142
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.6
0.780822
0.231579
0.168421
0.431486
0.389834
0.696659
1
0.790476
0
0.266667
1
0.55
0.452844
25.243206
0.064869
0.015797
25.162472
247,018
145,473
392,491
2,352.552381
1,385.457143
3,738.009524
52,275
4,472
497.857143
42.590476
0.6142
0.00585
105
{ "api": 105 }
46.206897
{ "mc": { "n": 55, "mean_similarity": 0.6991109091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.5636363636, "mean_accuracy_attempt...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
10
20260827_232625
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.663158
0.926471
0.284211
0.052632
0.433031
0.30114
0.77625
1
0.790476
0
0.504762
1
0.583333
0.602439
20.502301
1.242938
0.427367
18.831903
162,911
103,879
266,790
1,551.533333
989.32381
2,540.857143
38,914
4,142
370.609524
39.447619
0.462172
0.004402
105
{ "api": 105 }
57.241379
{ "mc": { "n": 55, "mean_similarity": 0.7615618182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4, "mean_coverage_ratio": 1, "accuracy": 0.5636363636, "mean_accuracy_attempted": 0.88...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
10
20260827_225743
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.631579
0.882353
0.284211
0.084211
0.486667
0.474323
0.848995
1
0.809524
0
0.514286
1
0.608333
0.645852
26.83254
1.072803
0.404211
25.355432
162,012
118,345
280,357
1,542.971429
1,127.095238
2,670.066667
38,868
4,620
370.171429
44
0.45833
0.004365
105
{ "api": 105 }
55.172414
{ "mc": { "n": 55, "mean_similarity": 0.8427509091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4363636364, "mean_coverage_ratio": 1, "accuracy": 0.5818181818, "mean_accuracy_attempted": 0.84...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
10
20260827_233655
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.747368
0.845238
0.115789
0.136842
0.647515
0.416942
0.745803
1
0.838095
0
0.590476
1
0.657143
0.630174
21.951044
2.024878
0.972395
18.953667
165,542
104,536
270,078
1,576.590476
995.580952
2,572.171429
44,070
5,537
419.714286
52.733333
0.430262
0.004098
105
{ "api": 105 }
59.310345
{ "mc": { "n": 55, "mean_similarity": 0.7336218182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8545454545, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.5454545455, "mean_coverage_ratio": 1, "accuracy": 0.7454545455, "mean_accuracy_attempt...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
10
20260827_234204
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.442105
0.736842
0.4
0.157895
0.397835
0.338367
0.737276
1
0.714286
0
0.171429
1
0.405714
0.526795
23.149007
0.087579
0.066805
22.994525
585,003
119,532
704,535
5,571.457143
1,138.4
6,709.857143
42,837
4,240
407.971429
40.380952
0.716115
0.00682
105
{ "api": 105 }
35.862069
{ "mc": { "n": 55, "mean_similarity": 0.73744, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2363636364, "mean_coverage_ratio": 1, "accuracy": 0.4909090909, "mean_accuracy_attempted": ...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
10
20260827_234653
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.589474
0.788732
0.252632
0.157895
0.517937
0.362132
0.761597
1
0.809524
0
0.028571
1
0.71619
0.657777
23.246089
0.095168
0.067125
23.083701
158,741
118,545
277,286
1,511.819048
1,129
2,640.819048
37,667
4,416
358.733333
42.057143
0.458219
0.004364
105
{ "api": 105 }
52.413793
{ "mc": { "n": 55, "mean_similarity": 0.7437745455, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7454545455, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0363636364, "mean_coverage_ratio": 1, "accuracy": 0.5636363636, "mean_accuracy_attempt...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
10
20260827_235205
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.442105
0.75
0.410526
0.147368
0.49537
0.396289
0.685539
1
0.742857
0
0.180952
1
0.593095
0.572163
23.942521
0.067219
0.070625
23.804575
247,654
127,134
374,788
2,358.609524
1,210.8
3,569.409524
57,288
5,296
545.6
50.438095
0.533741
0.005083
105
{ "api": 105 }
39.310345
{ "mc": { "n": 55, "mean_similarity": 0.68856, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7090909091, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.4727272727, "mean_accuracy_attempted": ...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
10
20260827_235744
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.673684
0.831169
0.189474
0.136842
0.636459
0.446868
0.774981
1
0.8
0
0.561905
1
0.65
0.643313
23.669789
1.416157
0.75138
21.502164
162,764
114,624
277,388
1,550.133333
1,091.657143
2,641.790476
42,431
5,713
404.104762
54.409524
0.47449
0.004519
105
{ "api": 105 }
57.931034
{ "mc": { "n": 55, "mean_similarity": 0.76274, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4909090909, "mean_coverage_ratio": 1, "accuracy": 0.6363636364, "mean_accuracy_attempted": ...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
10
20260827_233159
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.652632
0.849315
0.231579
0.115789
0.668214
0.604148
0.845741
1
0.790476
0
0.542857
1
0.775
0.641422
23.331442
1.031054
1.127135
21.17315
166,102
121,107
287,209
1,581.92381
1,153.4
2,735.32381
45,298
6,585
431.409524
62.714286
0.465962
0.004438
105
{ "api": 105 }
56.551724
{ "mc": { "n": 55, "mean_similarity": 0.8396581818, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7454545455, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4727272727, "mean_coverage_ratio": 1, "accuracy": 0.5818181818, "mean_accuracy_attempt...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
1
20260828_091405
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.694737
0.929577
0.252632
0.052632
0.543254
0.399056
0.760256
1
0.828571
0
0.571429
1
0.6
0.481281
25.10183
1.780485
0
23.321269
155,887
105,563
261,450
1,484.638095
1,005.361905
2,490
36,528
4,996
347.885714
47.580952
0.557311
0.005308
105
{ "api": 105 }
52.413793
{ "mc": { "n": 55, "mean_similarity": 0.7510054545, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4909090909, "mean_coverage_ratio": 1, "accuracy": 0.6, "mean_accuracy_attempted": 0.89...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
1
20260828_091848
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.736842
0.909091
0.189474
0.073684
0.541772
0.464957
0.786711
1
0.857143
0
0.361905
1
0.621905
0.645669
18.930006
0.095716
0
18.834225
588,398
100,179
688,577
5,603.790476
954.085714
6,557.87619
56,766
4,958
540.628571
47.219048
0.988758
0.009417
105
{ "api": 105 }
62.068966
{ "mc": { "n": 55, "mean_similarity": 0.7705563636, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8363636364, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.3454545455, "mean_coverage_ratio": 1, "accuracy": 0.6909090909, "mean_accuracy_attempt...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
1
20260828_092301
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.6
0.863636
0.305263
0.094737
0.58668
0.394643
0.777727
1
0.847619
0
0.047619
1
0.548571
0.5929
19.044268
0.095015
0
18.949192
148,110
92,915
241,025
1,410.571429
884.904762
2,295.47619
37,753
4,398
359.552381
41.885714
0.516727
0.004921
105
{ "api": 105 }
53.103448
{ "mc": { "n": 55, "mean_similarity": 0.7611909091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8181818182, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0727272727, "mean_coverage_ratio": 1, "accuracy": 0.4909090909, "mean_accuracy_attempt...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
1
20260828_092717
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.578947
0.820896
0.294737
0.126316
0.541349
0.401568
0.714385
1
0.742857
0
0.247619
1
0.561667
0.55849
19.574588
0.066361
0
19.508139
241,180
109,820
351,000
2,296.952381
1,045.904762
3,342.857143
51,206
5,136
487.67619
48.914286
0.682307
0.006498
105
{ "api": 105 }
48.275862
{ "mc": { "n": 55, "mean_similarity": 0.71458, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7272727273, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2545454545, "mean_coverage_ratio": 1, "accuracy": 0.5818181818, "mean_accuracy_attempted": ...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
1
20260828_093126
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.568421
0.9
0.368421
0.063158
0.463148
0.376368
0.789144
1
0.72381
0
0.457143
1
0.566667
0.636918
19.357182
1.212064
0
18.145023
150,440
89,437
239,877
1,432.761905
851.780952
2,284.542857
37,946
5,179
361.390476
49.32381
0.496742
0.004731
105
{ "api": 105 }
54.482759
{ "mc": { "n": 55, "mean_similarity": 0.7770309091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.6545454545, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.3636363636, "mean_coverage_ratio": 1, "accuracy": 0.4909090909, "mean_accuracy_attempt...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
no_reranker
null
null
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
1
20260828_090843
deepseek/deepseek-v4-pro-0813
openrouter
custom_list
105
0.578947
0.916667
0.368421
0.052632
0.547685
0.488115
0.854764
1
0.771429
0
0.47619
1
0.608333
0.577865
21.862068
0.998561
0
20.863429
162,061
94,470
256,531
1,543.438095
899.714286
2,443.152381
37,834
5,064
360.32381
48.228571
0.532657
0.005073
105
{ "api": 105 }
48.275862
{ "mc": { "n": 55, "mean_similarity": 0.8493709091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4, "mean_coverage_ratio": 1, "accuracy": 0.5090909091, "mean_accuracy_attempted": 0.87...
deepseek/deepseek-v4-pro-0813
DeepSeek V4 Pro 0813 (OpenRouter)
baseline
null
null
3gpp_baseline_deepseek-deepseek-v4-pro-0813_no_retrieval_final_3gpp_105
3gpp
baseline-deepseek-deepseek-v4-pro-0813
0
20260828_000540
final_3gpp_105.json
105
0.242105
0.244681
0.010526
0.747368
0
0.246294
0
0
-1
0
-1
-1
0
0.183752
33.011243
0
0
33.011219
26,578
216,711
243,289
253.12381
2,063.914286
2,317.038095
2,091
368
19.914286
3.504762
0.571124
0.005439
105
{ "api": 105 }
19.310345
{ "mc": { "n": 55, "mean_similarity": 0, "mean_distinct_sources_retrieved": 0, "canonical_hit_rate": -1, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": -1, "mean_coverage_ratio": -1, "accuracy": 0.2545454545, "mean_accuracy_attempted": 0.2592592593, "absten...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
10
20260828_001442
z-ai/glm-5.3
openrouter
custom_list
105
0.726316
0.811765
0.105263
0.168421
0.333558
0.223165
0.74511
1
0.87619
0
0.571429
1
0.676667
0.451973
19.576303
1.866977
0.315316
17.393885
145,890
143,971
289,861
1,389.428571
1,371.152381
2,760.580952
37,909
4,089
361.038095
38.942857
0.696176
0.00663
105
{ "api": 105 }
57.931034
{ "mc": { "n": 55, "mean_similarity": 0.7298509091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8909090909, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.5454545455, "mean_coverage_ratio": 1, "accuracy": 0.6909090909, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
10
20260828_001922
z-ai/glm-5.3
openrouter
custom_list
105
0.568421
0.683544
0.168421
0.263158
0.657579
0.517252
0.750733
1
0.752381
0
0.228571
1
0.512143
0.584674
20.150219
0.09228
0.022646
20.035187
388,644
176,489
565,133
3,701.371429
1,680.847619
5,382.219048
68,954
7,766
656.704762
73.961905
1.030856
0.009818
105
{ "api": 105 }
44.137931
{ "mc": { "n": 55, "mean_similarity": 0.7408618182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.4909090909, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
10
20260828_002329
z-ai/glm-5.3
openrouter
custom_list
105
0.610526
0.716049
0.147368
0.242105
0.570582
0.401359
0.759576
1
0.838095
0
0.066667
1
0.545
0.551456
18.952941
0.086997
0.013866
18.851959
139,425
158,582
298,007
1,327.857143
1,510.304762
2,838.161905
45,276
5,849
431.2
55.704762
0.793147
0.007554
105
{ "api": 105 }
50.344828
{ "mc": { "n": 55, "mean_similarity": 0.7469690909, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8363636364, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0909090909, "mean_coverage_ratio": 1, "accuracy": 0.6, "mean_accuracy_attempted": 0.71...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
10
20260828_002746
z-ai/glm-5.3
openrouter
custom_list
105
0.578947
0.670732
0.136842
0.284211
0.562694
0.412292
0.696663
1
0.790476
0
0.266667
1
0.55
0.447507
20.288959
0.065842
0.015019
20.207988
230,514
171,938
402,452
2,195.371429
1,637.504762
3,832.87619
60,270
5,945
574
56.619048
0.910271
0.008669
105
{ "api": 105 }
48.275862
{ "mc": { "n": 55, "mean_similarity": 0.6991163636, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.5272727273, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
10
20260828_003207
z-ai/glm-5.3
openrouter
custom_list
105
0.631579
0.789474
0.2
0.168421
0.480212
0.339865
0.77625
1
0.790476
0
0.504762
1
0.583333
0.589713
20.493606
1.297571
0.401459
18.794449
147,102
146,896
293,998
1,400.971429
1,399.009524
2,799.980952
48,536
5,279
462.247619
50.27619
0.708919
0.006752
105
{ "api": 105 }
55.172414
{ "mc": { "n": 55, "mean_similarity": 0.7615618182, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4, "mean_coverage_ratio": 1, "accuracy": 0.5636363636, "mean_accuracy_attempted": 0.77...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
10
20260828_001039
z-ai/glm-5.3
openrouter
custom_list
105
0.642105
0.73494
0.126316
0.231579
0.62045
0.565495
0.848995
1
0.809524
0
0.514286
1
0.608333
0.6445
20.263327
1.088369
0.390885
18.78395
145,311
156,464
301,775
1,383.914286
1,490.133333
2,874.047619
52,393
7,319
498.980952
69.704762
0.788055
0.007505
105
{ "api": 105 }
59.310345
{ "mc": { "n": 55, "mean_similarity": 0.8427509091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4363636364, "mean_coverage_ratio": 1, "accuracy": 0.6, "mean_accuracy_attempted": 0.66, "ab...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
10
20260828_004141
z-ai/glm-5.3
openrouter
custom_list
105
0.736842
0.777778
0.052632
0.210526
0.558034
0.394935
0.745797
1
0.838095
0
0.590476
1
0.657143
0.646593
22.027323
1.878852
0.97321
19.17514
151,878
148,572
300,450
1,446.457143
1,414.971429
2,861.428571
53,490
7,084
509.428571
67.466667
0.72349
0.00689
105
{ "api": 105 }
62.068966
{ "mc": { "n": 55, "mean_similarity": 0.7336109091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8545454545, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.5454545455, "mean_coverage_ratio": 1, "accuracy": 0.7272727273, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
10
20260828_004648
z-ai/glm-5.3
openrouter
custom_list
105
0.431579
0.554054
0.221053
0.347368
0.62244
0.455726
0.737269
1
0.714286
0
0.171429
1
0.405714
0.515174
22.708721
0.090341
0.069051
22.549204
567,030
189,577
756,607
5,400.285714
1,805.495238
7,205.780952
59,997
6,760
571.4
64.380952
1.204448
0.011471
105
{ "api": 105 }
35.172414
{ "mc": { "n": 55, "mean_similarity": 0.7374363636, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7636363636, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2363636364, "mean_coverage_ratio": 1, "accuracy": 0.4909090909, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
10
20260828_005120
z-ai/glm-5.3
openrouter
custom_list
105
0.642105
0.709302
0.094737
0.263158
0.617831
0.448062
0.761604
1
0.809524
0
0.028571
1
0.717857
0.656841
21.394364
0.092081
0.066944
21.235213
144,819
170,282
315,101
1,379.228571
1,621.733333
3,000.961905
46,939
6,179
447.038095
58.847619
0.847801
0.008074
105
{ "api": 105 }
62.758621
{ "mc": { "n": 55, "mean_similarity": 0.7437854545, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7454545455, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0363636364, "mean_coverage_ratio": 1, "accuracy": 0.6181818182, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
10
20260828_005615
z-ai/glm-5.3
openrouter
custom_list
105
0.452632
0.614286
0.263158
0.284211
0.661029
0.543845
0.685531
1
0.742857
0
0.180952
1
0.593095
0.623194
23.319587
0.068468
0.070916
23.180075
230,544
187,917
418,461
2,195.657143
1,789.685714
3,985.342857
76,780
9,488
731.238095
90.361905
0.981497
0.009348
105
{ "api": 105 }
46.896552
{ "mc": { "n": 55, "mean_similarity": 0.6885381818, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7090909091, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2181818182, "mean_coverage_ratio": 1, "accuracy": 0.4727272727, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
10
20260828_010100
z-ai/glm-5.3
openrouter
custom_list
105
0.673684
0.761905
0.115789
0.210526
0.587169
0.45507
0.774981
1
0.8
0
0.561905
1
0.65
0.642883
21.989936
1.131292
0.7511
20.107426
148,860
157,097
305,957
1,417.714286
1,496.161905
2,913.87619
53,747
7,629
511.87619
72.657143
0.793985
0.007562
105
{ "api": 105 }
61.37931
{ "mc": { "n": 55, "mean_similarity": 0.76274, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4909090909, "mean_coverage_ratio": 1, "accuracy": 0.6363636364, "mean_accuracy_attempted": ...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
grid
telecom-specific
otel-reranker-0.6b
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
10
20260828_003653
z-ai/glm-5.3
openrouter
custom_list
105
0.631579
0.705882
0.105263
0.263158
0.545582
0.517853
0.845741
1
0.790476
0
0.542857
1
0.775
0.677156
22.386389
1.115905
1.108558
20.161799
150,453
157,073
307,526
1,432.885714
1,495.933333
2,928.819048
50,772
6,437
483.542857
61.304762
0.770865
0.007342
105
{ "api": 105 }
58.62069
{ "mc": { "n": 55, "mean_similarity": 0.8396581818, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7454545455, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4727272727, "mean_coverage_ratio": 1, "accuracy": 0.5818181818, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
1
20260828_094419
z-ai/glm-5.3
openrouter
custom_list
105
0.663158
0.732558
0.094737
0.242105
0.485954
0.401463
0.760255
1
0.828571
0
0.571429
1
0.6
0.554297
24.505595
1.778208
0
22.727285
142,062
151,170
293,232
1,352.971429
1,439.714286
2,792.685714
49,232
7,055
468.87619
67.190476
0.739346
0.007041
105
{ "api": 105 }
57.241379
{ "mc": { "n": 55, "mean_similarity": 0.7510145455, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4909090909, "mean_coverage_ratio": 1, "accuracy": 0.6, "mean_accuracy_attempted": 0.66...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-0.6b_parent_child_llm_metadata_c512_o50
3gpp
otel-0.6b
1
20260828_094917
z-ai/glm-5.3
openrouter
custom_list
105
0.663158
0.732558
0.094737
0.242105
0.596355
0.464079
0.786719
1
0.857143
0
0.361905
1
0.621905
0.629435
20.358914
0.091064
0
20.267746
575,979
158,166
734,145
5,485.514286
1,506.342857
6,991.857143
65,982
6,799
628.4
64.752381
1.06994
0.01019
105
{ "api": 105 }
60.689655
{ "mc": { "n": 55, "mean_similarity": 0.7705654545, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8363636364, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.3454545455, "mean_coverage_ratio": 1, "accuracy": 0.6363636364, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-0.6b_parent_child_metadata_tagging_c512_o50
3gpp
otel-0.6b
1
20260828_095340
z-ai/glm-5.3
openrouter
custom_list
105
0.6
0.678571
0.115789
0.284211
0.625119
0.409481
0.777741
1
0.847619
0
0.047619
1
0.548571
0.638743
20.415648
0.092024
0
20.323531
136,746
157,720
294,466
1,302.342857
1,502.095238
2,804.438095
49,210
6,500
468.666667
61.904762
0.787354
0.007499
105
{ "api": 105 }
56.551724
{ "mc": { "n": 55, "mean_similarity": 0.7612054545, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8181818182, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.0727272727, "mean_coverage_ratio": 1, "accuracy": 0.5090909091, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-0.6b_sliding_window_tokens_none_c512_o50
3gpp
otel-0.6b
1
20260828_095836
z-ai/glm-5.3
openrouter
custom_list
105
0.547368
0.658228
0.168421
0.284211
0.545635
0.39836
0.714389
1
0.742857
0
0.247619
1
0.569286
0.543989
23.884752
0.068577
0
23.816061
228,039
183,213
411,252
2,171.8
1,744.885714
3,916.685714
64,466
6,594
613.961905
62.8
0.958022
0.009124
105
{ "api": 105 }
46.206897
{ "mc": { "n": 55, "mean_similarity": 0.7145854545, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7272727273, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.2545454545, "mean_coverage_ratio": 1, "accuracy": 0.5454545455, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
no_reranker
null
null
3gpp-r19_otel-109m_text_baseline_none_c1024_o200
3gpp
otel-109m
1
20260828_100308
z-ai/glm-5.3
openrouter
custom_list
105
0.547368
0.722222
0.242105
0.210526
0.365184
0.297561
0.789144
1
0.72381
0
0.457143
1
0.566667
0.576981
21.547469
1.135074
0
20.41227
138,690
149,700
288,390
1,320.857143
1,425.714286
2,746.571429
42,721
5,169
406.866667
49.228571
0.754642
0.007187
105
{ "api": 105 }
49.655172
{ "mc": { "n": 55, "mean_similarity": 0.7770309091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.6545454545, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.3636363636, "mean_coverage_ratio": 1, "accuracy": 0.4727272727, "mean_accuracy_attempt...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
no_reranker
null
null
3gpp-r19_bge_text_baseline_none_c1024_o200
3gpp
bge
1
20260828_093914
z-ai/glm-5.3
openrouter
custom_list
105
0.568421
0.710526
0.2
0.231579
0.411164
0.409295
0.854764
1
0.771429
0
0.47619
1
0.608333
0.49758
24.688159
1.015603
0
23.672434
150,942
158,795
309,737
1,437.542857
1,512.333333
2,949.87619
40,498
4,881
385.695238
46.485714
0.80189
0.007637
105
{ "api": 105 }
47.586207
{ "mc": { "n": 55, "mean_similarity": 0.8493709091, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.7818181818, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.4, "mean_coverage_ratio": 1, "accuracy": 0.5454545455, "mean_accuracy_attempted": 0.68...
z-ai/glm-5.3
GLM 5.3 (OpenRouter)
baseline
null
null
3gpp_baseline_z-ai-glm-5.3_no_retrieval_final_3gpp_105
3gpp
baseline-z-ai-glm-5.3
0
20260828_010344
final_3gpp_105.json
105
0.042105
0.042105
0
0.957895
0
0
0
0
-1
0
-1
-1
0
0
14.247575
0
0
14.247552
21,264
103,972
125,236
202.514286
990.209524
1,192.72381
0
0
0
0
0.487246
0.00464
105
{ "api": 105 }
2.758621
{ "mc": { "n": 55, "mean_similarity": 0, "mean_distinct_sources_retrieved": 0, "canonical_hit_rate": -1, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": -1, "mean_coverage_ratio": -1, "accuracy": 0.0727272727, "mean_accuracy_attempted": 0.0727272727, "absten...
qwen/qwen3.8-27b
Qwen3.8 27B (OpenRouter)
grid
generic-cross-encoder
ms-marco-minilm
3gpp-r19_otel-0.6b_text_baseline_none_c1024_o200
3gpp
otel-0.6b
10
20260828_011950
qwen/qwen3.8-27b
openrouter
custom_list
105
0.684211
0.844156
0.189474
0.126316
0.566136
0.337579
0.74511
1
0.87619
0
0.571429
1
0.676667
0.518745
36.360058
2.007902
0.284509
34.067556
170,801
167,806
338,607
1,626.67619
1,598.152381
3,224.828571
42,084
4,737
400.8
45.114286
0.562999
0.005362
105
{ "api": 105 }
55.172414
{ "mc": { "n": 55, "mean_similarity": 0.72984, "mean_distinct_sources_retrieved": 1, "canonical_hit_rate": 0.8909090909, "n_canonical_source_non_discriminating": 0, "exact_chunk_hit_rate": 0.5454545455, "mean_coverage_ratio": 1, "accuracy": 0.6727272727, "mean_accuracy_attempted": ...
End of preview.

telco-analysis

Aggregate (not per-question) evaluation results for my MSc thesis (GSMA Open Telco AI Initiative) — RAG configuration comparison across 6 generator models, built for supervisor review. Source data lives in LiamDuero/telco-eval; this repo is a curated, flattened extract of just the final numbers (accuracy, cost, latency, retrieval metrics, etc.) with the large per-question arrays stripped out.

Built by telcolens/liam/build_telco_analysis_repo.py, re-runnable at any time to refresh these files from the latest results on telco-eval.

Scope

Question set: final_3gpp_105.json (105 questions: 55 mc / 40 tf / 10 open, difficulty-first + series-stratified) — the one set all 6 models below share, so results are directly comparable.

Models (6):

Model Backend
google/gemma-4-31b-it local
KU-DFI/TelecomGPT-R1 local
deepseek/deepseek-v4-pro-0813 OpenRouter
z-ai/glm-5.3 OpenRouter
qwen/qwen3.8-27b OpenRouter
mistralai/mistral-small-2603 OpenRouter

Deliberately excluded: otel-2.0-local — confirmed broken (degenerate output: walls of tab/newline characters with a stray JSON fragment, same root cause as an earlier documented chat-template/tokenizer mismatch on that server, just a different garbage pattern) via a live content audit, not a real result.

Per model, up to 19 evaluation runs: 6 (chunking/enrichment/embedding) combos × 2 rerankers (generic cross-encoder + telecom-specific) + the same 6 combos with no reranker + 1 no-RAG baseline. Self-consistency: every RAG answer is 3 independent generations, majority-voted (mc/tf) or averaged (open).

Known gaps (present in the underlying data, not an error in this repo): Gemma has no baseline specifically against final_3gpp_105.json (only against the earlier small/medium sets); the 4 OpenRouter models were never run through the no-reranker leg (only Gemma and TelecomGPT-R1 were). all_results.json simply omits what doesn't exist — nothing here is fabricated or interpolated.

Files

  • all_results.json — every (model, phase, combo) result as one flat list. Each entry is a full aggregate result (accuracy, abstention/wrong rate, retrieval metrics, latency, tokens, real $ cost, by_format breakdown) tagged with model_id, phase (grid/no_reranker/baseline), reranker_label/reranker_key (null for no-reranker/baseline), and run_name.
  • mc_results.json / tf_results.json / open_results.json — the same runs, but only each one's per-format breakdown (by_format["mc"/"tf"/"open"]) pulled out into its own row, so a single format can be compared across every model/combo without digging through the nested by_format dict in all_results.json. mc/tf rows carry accuracy/abstention_rate/wrong_rate; open rows carry faithfulness/answer_relevancy/context_recall/answer_correctness. All rows also carry retrieval metrics (mean_similarity, canonical_hit_rate, etc.), since those apply regardless of format.

None of these files contain question_results (the per-question array with individual answers/ scores) — for that level of detail, pull the specific result.json from telco-eval directly.

Considerations

Personal working repo for thesis progress, not a curated public benchmark. Cost figures are real (OpenRouter's own billed usage.cost, or $0.00 for genuinely-free local vLLM), not estimates.

Downloads last month
115