Vision-language models (VLMs) face compositional safety risks where harmful intent emerges from the interaction between visual and textual inputs. As mixture-of-experts (MoE) VLMs become increasingly common, recent work has explored various safety interventions, including prompting, supervised fine-tuning, and routing-...
Zi-Yuan Yang, Wen-Xuan Ding, Shang-Bin Feng et al.· 0 citations
It is established that principled and informed team selection is critical and strong model selection algorithms for assembling effective multi-LLM systems are presented.
Zong-Wan Cao, Zi-Yuan Yang, Shang-Bin Feng et al.· 0 citations
A pool of language models can collaborate and improve collectively by learning from one another's responses. These interactions depend on the instructions used during training. Existing methods typically sample instructions uniformly, even though their usefulness may change as the models improve: an instruction on whic...
Christina Hahn, Shang-Bin Feng, Dean Light et al.· 0 citations
Contextual Early Reward (CER), which predicts terminal reward through behavioral evidence in a trajectory prefix, provides an interpretable, efficient, and dense evaluation method for long-horizon coding agents.
This work studies interventions at four levels of group-policy RL pipelines---data, rollout, reward, and advantage---and conducts extensive RL training on saturated reasoning data only, demonstrating that don't waste your saturated data: with the right strategies they can be recycled into useful RL training signals in...
FLIP (FLipped Inference for Prompt reconstruction), a reference-free and rubric-free reward modeling approach that reformulates reward modeling through backward inference that enables reliable reward modeling in downscaled regimes where judgment methods fail, is proposed.