FlashKDA ref impl lower bound set to -5 months ago
YouJiacheng · x · 2026-08-27
YulunDu notes that the FlashKDA reference implementation already had a default lower bound of -5 about four months ago, praising the impressive work from GLM.
More from Research
- Paper reveals why PPO value functions fail, proposes BPCO for stable training — heghbalz · 2026-08-27
- Gaussian fiddling brings facial expressions to Clug — repligate · 2026-08-27
- Lightwheel and Hugging Face release 100k-hour egocentric dataset for Physical AI — vanstriendaniel · 2026-08-27
- PyTorch Ecosystem Adds Perforated, TokenSpeed, and 8 Others — zhyncs42 · 2026-08-27
- V-Rubrics: Improving Visual Faithfulness via Rubric-Based Reinforcement Learning — liuziwei7 · 2026-08-27
- Interpretability research lacks a breakthrough moment; solving 'superintelligent psyche' may take much longer — ericjmichaud_ · 2026-08-27