US Labs Prefer RL for Reasoning, Chinese Labs Lean Towards SFT on Traces
xuanalogue · x · 2026-07-22
An observation highlights a divergence in reasoning training between US and Chinese AI labs: US labs favor Reinforcement Learning (RL), while Chinese labs more frequently use Supervised Fine-Tuning (SFT) on successful rollouts. The SFT approach is noted to be highly competitive, with Kimi K3 serving as soft evidence.
Related event: Divergent Sino-US Training Paradigms and New SFT Alignment Ideas(5 posts)→
More from Research
- EdgeRock open-sourced: 72 MiB KV cache at 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27
- EdgeRock-4B holds KV cache at 72 MiB across 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27
- TRL adds Score Centering, a simple fix for RLHF training instability — lvwerra · 2026-10-27
- Latent-Foresight: world models that learn to predict their own future — abursuc · 2026-10-27
- UK AISI papers: no good safeguards yet against misuse of biological AI tools — davidmanheim · 2026-10-27
- CMU's XYBench: LLMs answer queries literally and reinforce user misconceptions — valentina__py · 2026-10-27