cot_answer_extraction
Extract and normalize final answers from chain-of-thought outputs without scoring intermediate reasoning text.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- cot_answer_extraction
- description
- Extract and normalize final answers from chain-of-thought outputs without scoring intermediate reasoning text.
02 资源文件 2
03 SKILL.md 正文
Chain-of-Thought Answer Extraction
Use this skill when a model or recovery harness emits reasoning followed by a final answer and you need the shortest canonical answer span for scoring. Do not use it to build prompts, judge rationale quality, or run an external calculator.
Inputs
- Raw output text.
- Task type:
numeric,multiple_choice,yes_no,date,symbolic, orplan. - Optional final-answer markers and numeric tolerance settings.
Outputs
- Raw text.
- Extracted final answer.
- Normalized answer.
- Diagnostics with marker and fallback information.
Workflow
- Search from the end of the output for markers such as
The answer is,So the answer is, orAnswer:. - Extract the following short span and ignore earlier intermediate numbers.
- If no marker appears, fall back to a task-specific extraction from the end of the text.
- Normalize according to task type.
- Preserve diagnostics so recovery can distinguish explicit marker extraction from fallback behavior.
Validation
Run:
python scripts/cot_answer_extraction.py --self-test
python -m unittest discover -s tests
Tests cover numeric traces, multiple-choice letters, yes/no outputs, dates, and symbolic answers.
Limitations
The extractor intentionally keeps normalization conservative. It does not infer answers from unsupported world knowledge and it does not decide whether the reasoning path is faithful.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).