eval-harness

dataset

summeval_100

100 items · 41 arms. Headline: each arm's own primary metric, mean over its items; arms with different primary metrics are not ranked against each other.

baseline, promoted 2026-10-25

se_gemma_l_n100_v1

gemma_l (Gemma 4 31B, open weights, $0.0278 for 100 articles) is the cheapest judge the leader openai_l is not significantly different from. Check any judge against human judgements on your own outputs first. REPORT_SUMMEVAL, executive summary.

Inspect the 35 arms on tau, scored against gold/summeval_100Inspect the 3 arms on pair_agreement, scored against gold/summeval_100Inspect the 3 arms on top1, scored against gold/summeval_100Open its sheet

goals: A scorer that agrees with experts on summaries (SummEval)

01arms

scored against silverscored by a judgewrote the reference it is scored againstscored against a label set, not a reference setunmarked rows are scored against gold

41 shown
armmetricmeanitemscost per itemspeed per itemrunsscored against
se_gemma_l_pubtext_n100_v1 · pair
run on omarchy · Linux/x86_64, 32 CPUs
tau0.551100$0.001013 · 2,235 tokp50 2,417 · p95 4,134 ms1gold/summeval_100 goldinspect
se_openai_l_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.540100$0.006350 · 2,106 tokp50 2,205 · p95 2,643 ms1gold/summeval_100 goldinspect
se_gemma_l_pubtextcw_n100_v1 · pair
run on omarchy · Linux/x86_64, 32 CPUs
tau0.540100$0.000276 · 2,241 tokp50 2,639 · p95 6,340 ms1gold/summeval_100 goldinspect
se_gemma_l_n100_v1 baseline
run on omarchy · Linux/x86_64, 32 CPUs
tau0.538100$0.000278 · 2,251 tokp50 6,204 · p95 10,768 ms1gold/summeval_100 goldinspect
se_anthropic_l_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.528100$0.020077 · 3,151 tokp50 3,378 · p95 3,975 ms1gold/summeval_100 goldinspect
se_openai_m_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.528100$0.014153 · 2,106 tokp50 2,667 · p95 3,133 ms1gold/summeval_100 goldinspect
se_qwen37max_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.520100$0.004436 · 2,261 tokp50 4,416 · p95 5,240 ms1gold/summeval_100 goldinspect
se_qwen_s_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.519100$0.000385 · 2,297 tokp50 3,672 · p95 4,710 ms1gold/summeval_100 goldinspect
se_glm_m_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.517100$0.001144 · 2,198 tokp50 3,360 · p95 4,908 ms1gold/summeval_100 goldinspect
se_qwen_m_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.507100$0.001084 · 2,386 tokp50 7,054 · p95 7,789 ms1gold/summeval_100 goldinspect
se_glm_l_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.492100$0.002451 · 2,201 tokp50 3,130 · p95 6,118 ms1gold/summeval_100 goldinspect
se_deepseek_m_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.486100$0.000771 · 2,213 tokp50 877 · p95 2,130 ms1gold/summeval_100 goldinspect
se_anthropic_s_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.484100$0.003237 · 2,333 tokp50 2,315 · p95 3,192 ms1gold/summeval_100 goldinspect
se_anthropic_m_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.480100$0.008048 · 3,153 tokp50 2,966 · p95 3,359 ms1gold/summeval_100 goldinspect
se_gemma_m_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.477100$0.000252 · 2,247 tokp50 4,583 · p95 6,532 ms1gold/summeval_100 goldinspect
se_deepseek_l_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.469100$0.000258 · 2,220 tokp50 5,028 · p95 6,637 ms1gold/summeval_100 goldinspect
se_mistral_m_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.468100$0.001191 · 2,293 tokp50 1,836 · p95 2,678 ms1gold/summeval_100 goldinspect
se_deepseek_s_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.459100$0.000307 · 2,217 tokp50 2,535 · p95 9,703 ms1gold/summeval_100 goldinspect
se_mistral_l_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.455100$0.001257 · 2,281 tokp50 3,072 · p95 50,068 ms1gold/summeval_100 goldinspect
se_llama_l_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.454100$0.000594 · 2,250 tokp50 11,823 · p95 29,054 ms1gold/summeval_100 goldinspect
se_glm_s_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.446100$0.000481 · 2,282 tokp50 5,649 · p95 7,549 ms1gold/summeval_100 goldinspect
se_mistral_s_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.427100$0.000283 · 2,435 tokp50 8,940 · p95 14,768 ms1gold/summeval_100 goldinspect
se_gemma_s_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.420100$0.000279 · 2,329 tokp50 9,191 · p95 25,300 ms1gold/summeval_100 goldinspect
se_openai_s_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.411100$0.002124 · 2,106 tokp50 1,373 · p95 1,834 ms1gold/summeval_100 goldinspect
se_llama_m_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.406100$0.000517 · 2,209 tokp50 2,971 · p95 13,790 ms1gold/summeval_100 goldinspect
se_unievalpub_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.372100$0.000000p50 47,259 · p95 63,387 ms1gold/summeval_100 goldinspect
se_llama_s_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.353100$0.000297 · 2,313 tokp50 6,512 · p95 9,588 ms1gold/summeval_100 goldinspect
se_unieval_n100_v1
run on omarchy · Linux/x86_64, 32 CPUs
tau0.348100$0.000000p50 42,677 · p95 60,322 ms1gold/summeval_100 goldinspect
se_bartscore_n100_v1
run on Linux/x86_64, 4 CPUs
tau0.256100$0.000000p50 16,061 · p95 23,628 ms1gold/summeval_100 goldinspect
se_rouge2_n100_v1
run on Linux/x86_64, 4 CPUs
tau0.158100$0.000000p50 174 · p95 239 ms1gold/summeval_100 goldinspect
se_rouge1_n100_v1
run on Linux/x86_64, 4 CPUs
tau0.149100$0.000000p50 177 · p95 281 ms1gold/summeval_100 goldinspect
se_rougeL_n100_v1
run on Linux/x86_64, 4 CPUs
tau0.143100$0.000000p50 258 · p95 434 ms1gold/summeval_100 goldinspect
se_coverage_n100_v1
run on Linux/x86_64, 4 CPUs
tau0.101100$0.000000p50 877 · p95 1,128 ms1gold/summeval_100 goldinspect
se_length_n100_v1
run on Linux/x86_64, 4 CPUs
tau0.039100$0.000000p50 0.18 · p95 0.3 ms1gold/summeval_100 goldinspect
se_random_n100_v1
run on Linux/x86_64, 4 CPUs
tau0.003100$0.000000p50 0.12 · p95 0.21 ms1gold/summeval_100 goldinspect
se_gemma_l_n100_pairwise_v1
run on mac-dev · Linux/x86_64, 32 CPUs
pair_agreement0.825100$0.000764 · 7,503 tokp50 5,979 · p95 10,637 ms1gold/summeval_100 goldinspect
se_qwen_s_n100_pairwise_v1
run on mac-dev · Linux/x86_64, 32 CPUs
pair_agreement0.812100$0.000531 · 7,669 tokp50 11,556 · p95 59,941 ms1gold/summeval_100 goldinspect
se_random_n100_pairwise_v1
run on Linux/x86_64, 32 CPUs
pair_agreement0.473100$0.000000p50 0.064 · p95 0.12 ms1gold/summeval_100 goldinspect
se_gemma_l_n100_pick_v1
run on mac-dev · Linux/x86_64, 32 CPUs
top10.364100$0.000209 · 2,063 tokp50 1,124 · p95 7,045 ms1gold/summeval_100 goldinspect
se_qwen_s_n100_pick_v1
run on mac-dev · Linux/x86_64, 32 CPUs
top10.263100$0.000282 · 2,090 tokp50 1,283 · p95 2,276 ms1gold/summeval_100 goldinspect
se_random_n100_pick_v1
run on Linux/x86_64, 32 CPUs
top10.070100$0.000000p50 0.037 · p95 0.063 ms1gold/summeval_100 goldinspect

02what's inside

100 items, json, 1276–2344 words (median 1813, middle half 1620–2000).

Source: mteb/summeval (test split). The text stays there; each item below links to its row.

010201276–1382 words: 61383–1489 words: 71490–1596 words: 101597–1703 words: 171704–1810 words: 91811–1917 words: 141918–2024 words: 182025–2131 words: 122132–2238 words: 42239–2345 words: 31276–13821383–14891490–15961597–17031704–18101811–19171918–20242025–21312132–22382239–2345

the gold

gold/summeval_100: a record per item: coherence, consistency, fluency, relevance

100 shown
itemwordsgoldopen
cnn-test-40417144 fieldssource ↗ · inspect
cnn-test-47617004 fieldssource ↗ · inspect
cnn-test-51315814 fieldssource ↗ · inspect
cnn-test-88c15364 fieldssource ↗ · inspect
cnn-test-a0213624 fieldssource ↗ · inspect
cnn-test-b1c17794 fieldssource ↗ · inspect
cnn-test-b8b16304 fieldssource ↗ · inspect
cnn-test-bcd13204 fieldssource ↗ · inspect
cnn-test-c0516204 fieldssource ↗ · inspect
cnn-test-d7514774 fieldssource ↗ · inspect
cnn-test-e2b18334 fieldssource ↗ · inspect
cnn-test-e4918094 fieldssource ↗ · inspect
cnn-test-fbb14884 fieldssource ↗ · inspect
dm-test-02c916114 fieldssource ↗ · inspect
dm-test-03e218074 fieldssource ↗ · inspect
dm-test-051d23444 fieldssource ↗ · inspect
dm-test-078315414 fieldssource ↗ · inspect
dm-test-096e16294 fieldssource ↗ · inspect
dm-test-0cd121244 fieldssource ↗ · inspect
dm-test-0cdd16914 fieldssource ↗ · inspect
dm-test-0f0719944 fieldssource ↗ · inspect
dm-test-130f14204 fieldssource ↗ · inspect
dm-test-14c819844 fieldssource ↗ · inspect
dm-test-155716324 fieldssource ↗ · inspect
dm-test-174718814 fieldssource ↗ · inspect
dm-test-174919934 fieldssource ↗ · inspect
dm-test-182419314 fieldssource ↗ · inspect
dm-test-207d21564 fieldssource ↗ · inspect
dm-test-237720514 fieldssource ↗ · inspect
dm-test-23e814554 fieldssource ↗ · inspect
dm-test-26e416364 fieldssource ↗ · inspect
dm-test-2a4816964 fieldssource ↗ · inspect
dm-test-2abe16044 fieldssource ↗ · inspect
dm-test-2c3721084 fieldssource ↗ · inspect
dm-test-2c6621384 fieldssource ↗ · inspect
dm-test-2cf815404 fieldssource ↗ · inspect
dm-test-2cfc18594 fieldssource ↗ · inspect
dm-test-2fec18134 fieldssource ↗ · inspect
dm-test-311118744 fieldssource ↗ · inspect
dm-test-363518404 fieldssource ↗ · inspect
dm-test-378518184 fieldssource ↗ · inspect
dm-test-3ba319404 fieldssource ↗ · inspect
dm-test-400115674 fieldssource ↗ · inspect
dm-test-466013424 fieldssource ↗ · inspect
dm-test-4a5917704 fieldssource ↗ · inspect
dm-test-4e0d18814 fieldssource ↗ · inspect
dm-test-4f3b20454 fieldssource ↗ · inspect
dm-test-4fb619874 fieldssource ↗ · inspect
dm-test-54b818344 fieldssource ↗ · inspect
dm-test-5be020504 fieldssource ↗ · inspect
dm-test-632c13884 fieldssource ↗ · inspect
dm-test-637220174 fieldssource ↗ · inspect
dm-test-672119894 fieldssource ↗ · inspect
dm-test-6c1316474 fieldssource ↗ · inspect
dm-test-6f0216544 fieldssource ↗ · inspect
dm-test-7cba15004 fieldssource ↗ · inspect
dm-test-801715494 fieldssource ↗ · inspect
dm-test-808a17594 fieldssource ↗ · inspect
dm-test-81bd19724 fieldssource ↗ · inspect
dm-test-876416694 fieldssource ↗ · inspect
dm-test-884720924 fieldssource ↗ · inspect
dm-test-8b5118114 fieldssource ↗ · inspect
dm-test-8b5914704 fieldssource ↗ · inspect
dm-test-8bc816334 fieldssource ↗ · inspect
dm-test-8e6c22504 fieldssource ↗ · inspect
dm-test-97bb20004 fieldssource ↗ · inspect
dm-test-998c13534 fieldssource ↗ · inspect
dm-test-9afd20244 fieldssource ↗ · inspect
dm-test-9b1a20084 fieldssource ↗ · inspect
dm-test-9b5a20124 fieldssource ↗ · inspect
dm-test-9bea20084 fieldssource ↗ · inspect
dm-test-9d0f16844 fieldssource ↗ · inspect
dm-test-a3b620594 fieldssource ↗ · inspect
dm-test-b1f620694 fieldssource ↗ · inspect
dm-test-b25521674 fieldssource ↗ · inspect
dm-test-b59219724 fieldssource ↗ · inspect
dm-test-b5bc18794 fieldssource ↗ · inspect
dm-test-bd4614314 fieldssource ↗ · inspect
dm-test-c16c18164 fieldssource ↗ · inspect
dm-test-c50d19124 fieldssource ↗ · inspect
dm-test-c66617474 fieldssource ↗ · inspect
dm-test-cf4b16464 fieldssource ↗ · inspect
dm-test-d29621064 fieldssource ↗ · inspect
dm-test-d3c917844 fieldssource ↗ · inspect
dm-test-d68019964 fieldssource ↗ · inspect
dm-test-d83b19554 fieldssource ↗ · inspect
dm-test-d89d17184 fieldssource ↗ · inspect
dm-test-dea023224 fieldssource ↗ · inspect
dm-test-e06f15864 fieldssource ↗ · inspect
dm-test-e42814934 fieldssource ↗ · inspect
dm-test-e43e14934 fieldssource ↗ · inspect
dm-test-e47022224 fieldssource ↗ · inspect
dm-test-e7d820824 fieldssource ↗ · inspect
dm-test-e88020064 fieldssource ↗ · inspect
dm-test-eeef13214 fieldssource ↗ · inspect
dm-test-f26d18794 fieldssource ↗ · inspect
dm-test-f37f21074 fieldssource ↗ · inspect
dm-test-f46816474 fieldssource ↗ · inspect
dm-test-f5fe12764 fieldssource ↗ · inspect
dm-test-fada20624 fieldssource ↗ · inspect

Each row is the profile evh dataset-items --dataset-id summeval_100 prints; inspect opens the item in the inspector with the dataset's largest group of comparable arms, and its text where the inspector runs on a machine that has it.

03reference sets

04labels

No labels yet. A sheet to fill: evh label-sheet --dataset-id summeval_100 --out sheet.csv; a run's outputs as a model set: evh label-from-run --run R --set S.