arm
fn_deepseek_l_n200_v1
One config on few_nerd_280, scored one way; its repeats averaged per item.
- dataset
- few_nerd_280
- scored against
- gold/few_nerd_280 gold
- goal
- Named-entity extraction (Few-NERD)
- runs
- 1
- headline
- f1 0.587
- last run
- 2026-09-28 09:21
Inspect itAgainst the baselineWith the 28 arms scored the same way
01metrics
| metric | mean | items | better |
|---|---|---|---|
| f1 | 0.587 | 280 | higher |
| cost_usd | $0.000305 | 280 | lower |
| fn | 0.793 | 280 | neutral |
| fp | 1.461 | 280 | neutral |
| latency_ms | 3726.8 | 280 | lower |
| n_gold | 2.736 | 280 | neutral |
| n_predicted | 3.404 | 280 | neutral |
| parsed | 1.000 | 280 | higher |
| precision | 0.567 | 280 | higher |
| reasoning_tokens | 0.000 | 280 | neutral |
| recall | 0.639 | 280 | higher |
| tokens_in | 114.4 | 280 | lower |
| tokens_out | 65.1 | 280 | lower |
| tp | 1.943 | 280 | neutral |
| truncated | 0.000 | 280 | neutral |
| type_penalty | 0.107 | 280 | neutral |
| untyped_f1 | 0.694 | 280 | higher |
| untyped_fn | 0.454 | 280 | neutral |
| untyped_fp | 1.121 | 280 | neutral |
| untyped_tp | 2.282 | 280 | neutral |
02cost and speed
- cost per item
- $0.000305 · 180 tok
- speed per item
- p50 3,036 · p95 7,752 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 |
|---|---|---|---|---|
fn_deepseek_l_n200_v1_20260928T092110Z | 2026-09-28 09:21 | 280 | $0.085296 | rescore |
04config
config fn_deepseek_l_n200_v1 · adapter examples/ner-few-nerd/adapter.py · credentials are never shown
- family
deepseek- max_tokens
600- model
eval-deepseek-pro- price_tier
large- prompt_file
prompt.txt- provider
litellm- reasoning
{"enabled": false}- temperature
0.0- usd_per_mtok_in
0.28- usd_per_mtok_out
1.1
05conditions
- data.items_sha256
95a1b962707368cce0895143e73b63cd9884425502997ee1a93df99f97e2cad7- deployment.model
eval-deepseek-pro- deployment.source
adapter- fingerprint.version
5- library.datasets
5.0.1- library.numpy
1.26.4- library.openai
3.19.2- library.scikit-learn
1.9.1- 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
7cdd09fe0084ba29c616ab65fbfa08bf149e9006819641e9e91fe502ebe15a2b- scorer.id
examples/ner-few-nerd/adapter.py- scorer.sha256
d147da55107305786c21b75ac87d748b9ea4fb3e57ebdff622aafecd826559b1
06rounds
- Round r1 of Named-entity extraction (Few-NERD), closed: one of its arms.