Public KG record
Reinforcement Learning
Recorded overview
KG-recorded mechanism related to Reinforcement Learning; detailed English notes are not yet available.
An evidence-backed entity from the project knowledge graph; detailed tool metadata is still being added.
Recorded details
- Capabilities: 训练高频控制层
- Capabilities: audio-token merging
- Capabilities: temperature correction
- Performance: Voice of Reason 直接模型 GSM8K 达 70.3%
- Performance: Voice of Reason Stitch GSM8K 达 77.1%
- Performance: judge 与人工判断在 100 个案例中 88% 一致
- Requirements: 训练使用 16 张 H100 GPU
Evidence coverage
The KG currently records 9 known information dimensions and has 10 undocumented dimensions, linked to 26 public source articles. These counts describe record coverage, not model performance or verified fit.
Related public records
- AgentENV — applied to
- Molt — applied to
- Airbnb — uses
- alignment training — uses
- MuJoCo Warp — supports