3Blue1Brown (YouTube) · RSS 摘要

论文 Specification-first convergence with an AI coding agent: a case study of dismantling a core architectural invariant across 189 files in a 717k-line codebase with no test oracle and no human code review

攻略 Deep-Thinking Tokens：用"预测不稳定Token"衡量LLM真实推理努力 · 干货攻略

仓库 yifan1207/convergence-gap-instruction-tuning

论文 [Bespoke-Card] Why Tune When You Can Generate? Synthesizing Workload-Specific Cardinality Estimators

论文 The Galaxy's Guide to the Tokenizer: A Benchmark for Scientific Foundation Models

文档 Tom 评 flyP · 2026-07-09 · LongVQUBench 精读