works on

From the 2 of 19 linked papers with an AI index.

most citedDecoupled Alignment for Robust Plug-and-Play Adaptation

1 citations · 1 across the 3 of their papers we have counts for

collaborators

19 papers

cs.LG2026

Learn What's Left, Not What's Mastered: Saturation Aware Advantage Reweighting for Multi-Reward Policy Optimization

Yixuan Wang, Yifei Chen, Haichao Zhang +6

Reinforcement learning (RL) with group-relative advantages has become the de facto standard for post-training language model reasoners. However, when optimizing multiple reward obj…

cs.CL2026

Memory as a Controlled Process: Learned Adaptive Memory Management for LLM Agents

Eric Hanchen Jiang, Zhi Zhang, Yuchen Wu +11

The paper introduces MemCon, a framework that treats memory operations of large language model agents as a controllable Markov Decision Process, learning adaptive policies for when…

cs.CL20261 cited

Decoupled Alignment for Robust Plug-and-Play Adaptation

Haozheng Luo, Jiahao Yu, Wenxin Zhang +9

The paper proposes a training-free, plug-and-play method that uses knowledge distillation and model fusion to correct misaligned (shadow-aligned) large language models, improving s…

cs.CL2026

Compute Allocation in Evolutionary Search: From Depth-Breadth to Multi-Armed Bandits

Sixue Xing, Haoyu He, Kerui Wu +4

LLM-guided evolutionary search (Evolve systems) has reached state-of-the-art results on mathematical and combinatorial tasks, yet most existing systems report only the best of many…

cs.AI2026

Contrastive Reasoning Alignment: Reinforcement Learning from Hidden Representations

Haozheng Luo, Yimin Wang, Jiahao Yu +2

We propose CRAFT, a red-teaming alignment framework that leverages model reasoning capabilities and hidden representations to improve robustness against jailbreak attacks. Unlike p…

cs.LG2026

Attention Sinks and Outliers in Attention Residuals

Haozheng Luo, Haoran Dai, Shaoyang Zhang +10

We propose OASIS, an outlier- and sink-aware technique built on inter-layer null signaling. As AttnResidual architectures introduce an additional depth-wise normalization channel,…