Nathan Lambert 发布后训练最大开放问题讲义:RL 算力已超预训练
rickasaurus · x · 2026-11-22
Nathan Lambert released a slide deck on the biggest open problems in post-training and the field's evolution, outlining Trillium Labs' goals. Key research areas: harness specialization and environment design for context management, character training (model personality, safety, trust), and open-source RL environments. The deck ends with a debate on how RSI is changing research. Accompanying data shows RL compute share exploding: DeepSeek R1 used 5% as much RL as pretraining compute, while Beam (Oct 2026) spent >1.7× more on RL (7.1M vs <4.1M GPU-hours) and Inkling 2 Small 4×.
所属事件:Nathan Lambert 发后训练开放问题讲稿,称 RL 算力已超预训练(3 条相关)→
「模型」频道最新
- Gary Marcus 开呛:有人连 Claude Code 是神经符号系统都不知道 — mattyp · 2026-11-22
- Nemotron 3.6 Academia 即将登陆 llama.cpp:循环混合架构获原生支持 — maternion · 2026-11-22
- 爆料称 OpenAI 内部士气低落:GPT-6.2 口碑翻车,谷歌一周反超 — SumitGup · 2026-11-22
- Gemini 4 Neon 单任务成本约 30 美分,与 Fable 5.2 相当 — ChrisGPT · 2026-11-22
- 实测发现:AI 面对开放数学题常直接弃权,称「这题是未解难题」 — Jsevillamol · 2026-11-22
- 数学家回应 Anthropic RH 证明争议:用到所有工具反而是好兆头 — harris_edouard · 2026-11-22