arm
se_qwen_s_n100_pairwise_v1
One config on summeval_100, scored one way; its repeats averaged per item.
- dataset
- summeval_100
- scored against
- gold/summeval_100 gold
- runs
- 1
- headline
- pair_agreement 0.812
- last run
- 2026-11-02 16:01
01metrics
| metric | mean | items | better |
|---|---|---|---|
| pair_agreement | 0.812 | 100 | higher |
| cost_usd | $0.000531 | 100 | lower |
| latency_ms | 20081.2 | 100 | lower |
| n_pairs | 7.490 | 100 | neutral |
| parsed | 0.999 | 100 | neutral |
| reasoning_tokens | 0.000 | 100 | neutral |
| tokens_in | 7590.9 | 100 | lower |
| tokens_out | 77.7 | 100 | lower |
| truncated | 0.010 | 100 | neutral |
02cost and speed
- cost per item
- $0.000531 · 7,669 tok
- speed per item
- p50 11,556 · p95 59,941 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 |
|---|---|---|---|---|
se_qwen_s_n100_pairwise_v1_20261102T152804Z | 2026-11-02 16:01 | 100 | $0.053094 | measured |
04config
config se_qwen_s_n100_pairwise_v1 · adapter examples/summeval-judging/adapter.py · credentials are never shown
- family
qwen- max_tokens
1500- mode
pairwise- model
eval-qwen-flash- price_tier
small- prompt_file
prompt_pairwise.txt- provider
litellm- reasoning
{"enabled": false}- temperature
0.0- usd_per_mtok_in
0.05- usd_per_mtok_out
0.2
05conditions
- data.items_sha256
b59999b281b2cbe87c57e4af34cc475c01ba1d531c7b77cb93a136665119c613- deployment.model
eval-qwen-flash- deployment.source
adapter- deployment.upstream
openrouter/qwen/qwen3.8-flash- fingerprint.version
5- harness.code_sha256
fd4cd92c896bc9d1d43a63ce5809ee1a831a95a66ae786275a750501d93ac77a- harness.version
0.5.0- library.anthropic
2.3.0- library.numpy
2.5.4- library.openai
3.41.0- library.rouge-score
0.1.2- machine.cpu_count
32- machine.label
mac-dev- machine.platform
Linux/x86_64- machine.python
3.14.7- machine.release
7.2.5-3-omarchy- reference.sha256
7b5e618e23b7f2b222665d83e99006eb0e9a51066e63e2e5ff8c51ae3c696c21- scorer.id
examples/summeval-judging/adapter.py- scorer.sha256
2ad9193bcb91f7218b9eac0c9e9f68daf372e989cdfa2cd318e4833e693c88c9