Tencent Hunyuan releases FlowBalance, a verifier-grounded self-improvement method for reasoning models
Tencent-Hunyuan · hf · 2026-09-08
Tencent's Hunyuan team published FlowBalance, a research project for improving reasoning models through verifier-grounded self-improvement.
The method uses trajectory-level score reweighting and profile-based trajectory balance to calibrate self-guidance with a verifier, leveraging on-policy reasoning experience rather than external labels.
More from Research
- EmbodiedSkills: A Unified Framework for Orchestrating VLA Robot Agents — Wei Wang · 2026-09-08
- One Symptom, Three Levers: A Critical Review of On-Policy Self-Distillation — OVHaiLLM · 2026-09-08
- Fluid Regularity Landscape Moving Fast: Tao Says Smooth NS Now Looks Feasible — basedjensen · 2026-09-08
- CIRPIN trains equivariant GNNs on circular permutations to fix protein structure search blind spots — bravo_abad · 2026-09-08
- JapanFold launches: 8 open biology AI models served free for Japanese researchers — DavidBennett__ · 2026-09-08
- Anandkumar team proves stable singularity for 3D Euler using PINN-guided math proof — AnimaAnandkumar · 2026-09-08