references

참고 출처 — 05 환각·신뢰도

#	파일	인용 위치	verbatim
S1	sources/05-ai-fundamentals/llm-basics/rlhf-chip-huyen.md	01 — RLHF paradox	false
S2	sources/05-ai-fundamentals/llm-basics/llm-powered-autonomous-agents-lilian-weng.md	01·02 — agent limitations	false
S3	sources/14-research-notes/2026-04-07-anthropic-claude-code-best-practices.md	02·예시01 — explore→plan→code→commit, subagents	false
S4	sources/01-anthropic-official/api/tool-use-overview.md	02 — 도구 사용의 능력 향상	true
S5	sources/18-research-papers/eval/swe-bench-arxiv-2310-06770.md	02 — SWE-bench	true
S6	sources/18-research-papers/eval/swe-bench-verified-openai-2024.md	02 — SWE-bench Verified	true
S7	sources/18-research-papers/eval/livecodebench-arxiv-2403-07974.md	02 — LiveCodeBench, 학습 누수	true

verbatim:false 주의

S1·S2·S3는 요약본. 정확한 영문 표현은 원문 URL로 재확인.