arm
cnn_bart_l_n200_v1 baseline
One config on cnn_dailymail_200, scored one way; its repeats averaged per item.
- dataset
- cnn_dailymail_200
- scored against
- gold/cnn_dailymail_200 gold
- goal
- Short news summaries (CNN)
- runs
- 1
- headline
- coverage 0.346
- last run
- 2026-09-26 16:14
01metrics
| metric | mean | items | better |
|---|---|---|---|
| coverage | 0.346 | 200 | higher |
| compression | 0.132 | 200 | neutral |
| concision | 0.261 | 200 | higher |
| cost_usd | $0.000000 | 200 | lower |
| fmt_bullets | 0.000 | 200 | neutral |
| fmt_label | 0.000 | 200 | neutral |
| fmt_markdown | 0.000 | 200 | neutral |
| fmt_narration | 0.010 | 200 | neutral |
| fmt_overlong | 0.000 | 200 | neutral |
| fmt_paragraphs | 0.000 | 200 | neutral |
| grounding | 0.884 | 200 | higher |
| input_truncated | 0.230 | 200 | neutral |
| latency_ms | 11009.6 | 200 | lower |
| length_vs_reference | 1.579 | 200 | neutral |
| reference_words | 34.7 | 200 | neutral |
| rouge1 | 0.348 | 200 | higher |
| rougeLsum | 0.316 | 200 | higher |
| summary_words | 50.8 | 200 | neutral |
| tokens_in | 697.2 | 200 | lower |
| tokens_out | 66.2 | 200 | lower |
02cost and speed
- cost per item
- $0.000000 · 763 tok
- speed per item
- p50 10,724 · p95 14,371 ms
Per item, over its items: cost as dollars and tokens (a local model is free but still uses tokens), speed as the median and the slow tail (p95). A goal that names this arm weighs them in its decision.
03runs
| run | made | items | cost | how |
|---|---|---|---|---|
cnn_bart_l_n200_v1_20260926T153757Z | 2026-09-26 16:14 | 200 | $0.000000 | measured |
04config
config cnn_bart_l_n200_v1 · adapter examples/summarization-cnn-dailymail/adapter.py · credentials are never shown
- device
cpu- early_stopping
true- family
bart- length_penalty
2.0- max_length
142- min_length
56- model
facebook/bart-large-cnn- no_repeat_ngram_size
3- num_beams
4- precision
fp32- price_tier
free- provider
hf_local
05deploy sheet
- measured by
cnn_bart_l_n200_v1_20260926T153757Zon cnn_dailymail_200- against
- gold/cnn_dailymail_200
- model sent
facebook/bart-large-cnn- model record
facebook-bart-large-cnn- server
- not recorded
- build
- 37f520fa929c961707657b28798b30c003dd100b
- settings sent
device: cpuearly_stopping: truelength_penalty: 2.0max_length: 142min_length: 56model: facebook/bart-large-cnnno_repeat_ngram_size: 3num_beams: 4precision: fp32provider: hf_local- not sent
- bookkeeping:
family, price_tier - declared
deployment.weight_files: ['model.safetensors:1625222120']- reported per item
encoder_window: 1024- adapter
examples/summarization-cnn-dailymail/adapter.py51a7b06b1a2e- harness
- fb7366a0a694
No capture here says what production ran (the adapter's production_ran).
Command: evh deploy-sheet --config cnn_bart_l_n200_v1 [--capture C]
06conditions
- data.items_sha256
257feda902391b934b8cde940c15e91201096dbe96ec5a81337e9da007ce3ec7- deployment.build
37f520fa929c961707657b28798b30c003dd100b- deployment.device
cpu- deployment.model
facebook/bart-large-cnn- deployment.quantization
fp32- deployment.source
adapter- fingerprint.version
5- library.numpy
1.26.4- library.openai
3.19.2- library.rouge-score
0.1.2- library.torch
2.2.2- library.transformers
4.55.4- machine.cpu_count
12- machine.label
mac-dev- machine.platform
Darwin/x86_64- machine.processor
i386- machine.python
3.11.8- machine.release
24.6.0- reference.sha256
5d0b02c3798017535572d24dd3ce0592332938a0c91e93684d163963a60a7cd9- scorer.id
examples/summarization-cnn-dailymail/adapter.py- scorer.sha256
51a7b06b1a2eec722a3bd024b7ed832746d06ef44683dd0e9fa8abc5c18d5618
07the same config, scored other ways
scored against silverscored by a judgescored against a label set, not a reference setunmarked rows are scored against gold
| arm | scored against | headline | mean |
|---|---|---|---|
| cnn_bart_l_n200_v1 · linux | gold/cnn_dailymail_200 gold | coverage | 0.346 |
| cnn_bart_l_n200_v1 · silver | silver/cnn_dailymail_200 silver | coverage | 0.317 |
| cnn_bart_l_n200_v1 · judged/judge-summ-pick-v1 | judge judge-summ-pick-v1 | judge_won | 0.000 |
| cnn_bart_l_n200_v1 · judged/judge-summ-pointwise-v1 | judge judge-summ-pointwise-v1 | judge_score_unvalidated | 4.181 |
| cnn_bart_l_n200_v1 · judged/judge-summ-v1 | judge judge-summ-v1 | judge_score | 3.868 |
| cnn_bart_l_n200_v1 · judged/judge-summ-v2 | judge judge-summ-v2 | judge_score | 3.821 |
08rounds
- Round r1 of Short news summaries (CNN), closed: one of its arms.