FSPO：函数空间时间自蒸馏让 7B/8B 模型推理训练更快更省
IITK · hf · 2026-10-27
A new paper proposes Function-Space Policy Optimization (FSPO), combining temporal consolidation with privileged adaptation to improve reasoning and tool-use policy optimization:
- Results: on Qwen3-8B and Olmo3-7B-Instruct, FSPO reaches avg@16 scores of 76.2 and 72.4 across five tasks (chemistry/physics/biology/materials/tool use) within a 5-hour training budget, beating GRPO and SDPO baselines (e.g., Qwen3-8B: 76.2 vs GRPO's 70.4)
- Efficiency: achieves highest accuracy with the fewest average generated tokens and reduced memory use
- Experiments ran on 4 NVIDIA GH200 GPUs following the SDPO training setup
「研究」频道最新
- 研究者与 LLM 讨论论文后,让模型改写出 12 页更清晰版本(正确性未验证) — thomasahle · 2026-10-27
- OpenAI 宣布数学新成果，被指抢先公开社区数学家 YM4 工作 — burny_tech · 2026-10-27
- 纽约将办可靠 AI 见面会，首期聚焦合成数据 — karen_ullrich · 2026-10-27
- Meta 发布 InteractiveWorld：可验证环境测试 MLLM 能否当人类现场助手 — meta · 2026-10-27
- 小米 MiLM 联合武大推出 WorldCast：大角度运镜下角色视频保持世界一致 — Lei Cao · 2026-10-27
- UCL 发布持续学习基准 CLE：最强前沿模型仅通过 47.7% 关卡 — UniversityCollegeLondon · 2026-10-27