arm
cnn_llama_m_n200_v1 · judged/judge-summ-v2
One config on cnn_dailymail_200, scored one way; its repeats averaged per item.
- dataset
- cnn_dailymail_200
- scored against
- judge
judge-summ-v2 - goal
- Short news summaries (CNN)
- runs
- 1
- headline
- judge_score 4.488
- last run
- 2026-10-25 13:39
01metrics
| metric | mean | items | better |
|---|---|---|---|
| judge_score | 4.488 | 200 | higher |
| cost_usd | $0.000165 | 200 | lower |
| judge_anthropic_l | 4.436 | 200 | higher |
| judge_openai_l | 4.540 | 200 | higher |
| latency_ms | 4045.2 | 200 | lower |
| reasoning_tokens | 0.000 | 200 | neutral |
| tokens_in | 745.1 | 200 | lower |
| tokens_out | 63.8 | 200 | lower |
| truncated | 0.000 | 200 | neutral |
02cost and speed
- cost per item
- $0.000165 · 809 tok
- speed per item
- p50 2,682 · p95 11,240 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_llama_m_n200_v1_20260925T235110Z | 2026-10-25 13:39 | 200 | $0.032961 | judge |
04config
config cnn_llama_m_n200_v1 · adapter examples/summarization-cnn-dailymail/adapter.py · credentials are never shown
- family
llama- max_tokens
1200- model
eval-llama-70b- price_tier
mid- prompt_file
prompt.txt- provider
litellm- reasoning
{"enabled": false}- temperature
0.0- usd_per_mtok_in
0.12- usd_per_mtok_out
0.3
05conditions
- data.items_sha256
257feda902391b934b8cde940c15e91201096dbe96ec5a81337e9da007ce3ec7- deployment.model
eval-llama-70b- deployment.source
adapter- deployment.upstream
openrouter/meta-llama/llama-3.3-70b-instruct- fingerprint.version
5- library.numpy
2.4.6- library.openai
3.19.2- library.rouge-score
0.1.2- machine.cpu_count
12- machine.platform
Darwin/x86_64- machine.processor
i386- machine.python
3.11.8- machine.release
24.6.0- reference.sha256
5d0b02c3798017535572d24dd3ce0592332938a0c91e93684d163963a60a7cd9- scorer.id
judge-score:judge-summ-v2- scorer.sha256
4b5440bf0599fd12fd2d69a2697f1b0cdf54871486ef228b60be554750e82887
06the 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_llama_m_n200_v1 | gold/cnn_dailymail_200 gold | coverage | 0.338 |
| cnn_llama_m_n200_v1 · silver | silver/cnn_dailymail_200 silver | coverage | 0.358 |
| cnn_llama_m_n200_v1 · judged/judge-summ-pick-v1 | judge judge-summ-pick-v1 | judge_won | 0.001 |
| cnn_llama_m_n200_v1 · judged/judge-summ-v1 | judge judge-summ-v1 | judge_score | 4.674 |