IX2-0345
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Accuracy Evaluations - Line hint:
564
Context Before
Throughput optimizations can sometimes quietly trade off accuracy (e.g. via aggressively relaxed acceptance rates, decoding tweaks, numerically unstable kernels, or endpoint misconfiguration). Without evals, a misconfigured server (truncation, bad decoding, wrong endpoint params) can still produce great throughput numbers but deliver garbage answers. For example, this additional layer of checks has helped us discover issues with some DP attention implementation for GPT-OSS.
Each representative throughput config now has an associated numerical accuracy check. Currently we are only using GSM8k, but being a very easy benchmark, the evaluation scores may not change much from differences in numerical calculation, and a harder benchmark may have a larger delta with respect to numerical accuracy. Thus, we plan to expand towards harder ones in the future, such as GPQA, HLE, MATH-500, SWE-Bench verified.
Evidence
Another form of performance-accuracy tradeoff is quantization
Context After
Source: SemiAnalysis InferenceX ↗
② Atomic Claim
另一種 performance-accuracy trade-off 是 quantization。
- Epistemic Mode:
ASSERTED - Mapping Status:
PARTIAL
③ Semantic Frame
{
"attribute": "UNSPECIFIED_ATTRIBUTE",
"context_nodes": [],
"entity": {
"id": "04_knowledge_base/Quantization",
"label": "quantization"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [],
"value_text": "另一種 performance-accuracy trade-off 是 quantization。"
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| entity | quantization | Quantization |
⑤ Human Review
請在 Properties 逐項確認：
- 原文 → Atomic Claim 是否忠實
- Atomic Claim → Semantic Frame 是否忠實
- Canonical Entity mapping 是否正確
- Epistemic mode 是否保留原文語氣
- 最後選擇
review_action
Review state
Markdown 內文不是正式 approval。只有 Apply bridge 寫入的 Decision Ledger event 才是正式決策。