AERIS is an offline policy improvement framework for multi-UAV integrated sensing and communication (ISAC) that learns from fixed flight logs using centralized training and decentralized execution. It introduces STAR-CRDT, an offline multi-agent RL algorithm that rectifies local actions and distills trusted improvements into decentralized actors, providing an offline-support policy improvement guarantee. Experiments demonstrate that STAR-CRDT boosts the main ISAC objective return by 29.3% and improves communication sum rate, sensing pass rate, and sensing margin while reducing collision-risk events by 54.2%.
By Ziyuan Wang (Steven), Yifan Sui (Steven), Wei Wei (Steven), Wenjie Xin (Steven), Zekai Zhang (Steven), Xiangwang Hou (Steven), Xiao-Ping (Steven), Zhang
The paper investigates whether fine‑tuning a language model erases previously embedded activation steering interventions that suppress refusals and encourage brevity. Across five instruction‑tuned models (3B–14B) subjected to non‑adversarial supervised fine‑tuning (SFT) and reinforcement learning from human feedback (RLHF), the authors find that the steering’s behavioural effect degrades when the fine‑tuning objective conflicts with the targeted behaviour, yet the underlying weight edits remain largely unchanged. Mechanistically, the steering vectors survive with minimal alteration, but functionally the steering is vulnerable and must be re‑validated after downstream training.
By Philipp E. Glass, Allan Tucker, Yongmin Li, Alina Miron
The paper introduces PLUS, a framework that uses reinforcement learning to generate text-based summaries of individual users’ preferences, characteristics, and past conversations. These summaries condition a reward model, allowing it to predict personalized response preferences and improving reward accuracy by 11–77 % over the standard Bradley‑Terry model. PLUS demonstrates robust performance with new users and topics, achieves a 25 % improvement over existing personalized RLHF techniques, and enables zero‑shot personalization for state‑of‑the‑art models like GPT‑4.
By Hyunji Nam, Yanming Wan, Mickel Liu, Peter Ahnn, Jianxun Lian, Natasha Jaques
The paper investigates how different estimators of the reverse Kullback–Leibler (KL) divergence used as a regularization term in reinforcement learning (RL) training of large language models (LLMs) affect training stability and downstream performance. By analyzing gradient bias across various estimator configurations, the authors demonstrate that biased gradients can cause training instabilities, while unbiased configurations improve performance on both in‑domain and out‑of‑domain tasks. Experiments on Qwen2.5‑7B, Llama‑3.1‑8B‑Instruct, and Qwen3‑4B‑Instruct‑2507 confirm these findings and show that KL regularization also stabilizes off‑policy RL training in asynchronous setups.
By Vedant Shah, Johan Obando-Ceron, Vineet Jain, Brian Bartoldson, Bhavya Kailkhura, Sarthak Mittal, Glen Berseth, Pablo Samuel Castro, Yoshua Bengio, Esmeralda S. Whitammer, Moksh Jain, Siddarth Venkatraman, Aaron Courville
Emergent Abilities in Large Language Models: A Survey reviews how scaling LLMs leads to previously unseen capabilities such as advanced reasoning, in-context learning, coding, and problem-solving. The paper critically examines definitions, inconsistencies, and the conditions that foster these abilities, including scaling laws, task complexity, pre‑training loss, quantization, and prompting strategies. It also discusses the extension to Large Reasoning Models and highlights safety concerns like deception, manipulation, and reward hacking, calling for improved evaluation and governance.
By Leonardo Berti, Flavio Giorgi, Gjergji Kasneci
The paper introduces $R^3$, a post‑training method that converts vision‑language models into robotic reasoners by first mid‑training on expert reasoning traces and then refining them with single‑step rubric‑based reinforcement learning. $R^3$ enables free‑form language reasoning to guide low‑level manipulation policies, improving exploration, generalization, and performance on long‑horizon tasks in Language Table and simulated bimanual grocery packing benchmarks. The approach outperforms instruction‑only imitation learning baselines and demonstrates that natural language reasoning can serve as a test‑time compute mechanism for steering robotic actions.
By Lehong Wu, Yuxiao Qu, Zheyuan Hu, Ivan Zhang, Limin Wei, Zackory Erickson, Aviral Kumar
ResMerge is a new framework for merging large language models trained via reinforcement learning. It separates each model’s task vector into a leading spectral head and a residual component, finding that both parts contain valuable behavior knowledge but behave differently during merging. The method builds a stable residual backbone using Spherical Residual Consensus Adaptation and then adds a lightweight head correction module that activates only when experts agree, leading to better preservation of expert capabilities compared to existing merging baselines.
By Yandu Sun, Zhiyan Hou, Hongyan An, Weizhen Wang, Haokai Ma, Yuheng Jia, Junfeng Fang, Haiyun Guo, Jinqiao Wang
The paper introduces Golden-GRPO Injection (GRIN), a three-stage self‑learning framework that uses a mixed‑policy reinforcement learning algorithm to inject knowledge into large language models. GRIN injects a golden answer to provide learning signals even when on‑policy rollouts fail on novel facts, and is evaluated on two new document‑level benchmarks—Blank and Counter—that test novel acquisition and counterfactual overwrite. Experiments show that mixed‑policy RL enables knowledge absorption beyond what supervised fine‑tuning can achieve, with GRIN outperforming SFT and other RL baselines on harder question types while matching them on basic fact recall.
By Zhibo Hou, Fan Zhao, Zhiyu An, Wan Du
MMEmb-R1 is a multimodal embedding framework that enhances reasoning by treating it as a latent variable and selecting beneficial reasoning paths through pair-aware selection and counterfactual intervention. It uses reinforcement learning to invoke reasoning only when necessary, reducing unnecessary computation and latency. On the MMEB-V2 benchmark, MMEmb-R1 achieves a state‑of‑the‑art score of 71.2 with just 4 B parameters.
By Yuchi Wang, Dingkang Yang, Haiyang Yu, Weikang Bian, Jiefeng Long, Xiao Liang, Chao Feng, Hongsheng Li
VGA‑BenchV2 is an expanded, human‑aligned benchmark and optimization framework that jointly evaluates video generation quality and aesthetic value. It builds on the original VGA‑Bench taxonomy, adding 52 sub‑dimensions and 1,016 curated prompts to generate over 60,000 videos from 12 mainstream models. The benchmark significantly enlarges human supervision with 36,000 task‑level annotations and introduces a hybrid evaluator (VAQA‑Net, VTag‑Net, VGQA‑Net) that aligns well with human judgments and can be used as a reward model for reinforcement‑learning fine‑tuning.
By Longteng Jiang, DanDan Zheng, Qianqian Qiao, Heng Huang, Huaye Wang, Yihang Bo, Bao Peng, Jingdong Chen, Jun Zhou, Xin Jin
Reasoning in language allows foundation models to spend more test-time compute on hard problems, such as those requiring decomposition, constraint tracking, and prediction of future consequences. Whet...
Robust Markov Decision Processes (RMDPs) generalize classical MDPs by allowing uncertainty in transition probabilities and optimizing against their worst-case realization. We consider $(s,a)$-rectangu...
On-policy distillation trains a language model on its own generations while a teacher scores them token by token. It combines the dense supervision of imitation learning with the on-policy sampling of...
Large language models access knowledge inconsistently across languages, but to what extent do they differ in their skill sets when interacting with different languages? This work quantifies cross-ling...
We introduce VGA-BenchV2, an extended human-aligned benchmark and optimization framework for jointly evaluating and improving video generation quality and aesthetic value. Built upon VGA-Bench, VGA-Be...
arXiv:2608.23664v1 Announce Type: cross
Abstract: Reward fine-tuning is becoming an important tool for adapting diffusion models to human preferences and task-specific objectives, but existing method...
By Jaemoo Choi, Wei Guo, Yuchen Zhu, Arash Vahdat, Molei Tao, Julius Berner, Yongxin Chen
arXiv:2604.01024v2 Announce Type: replace
Abstract: We study model-based learning of finite-window policies in tabular partially observable Markov decision processes (POMDPs). A common approach to le...
By Philip Jordan, Maryam Kamgarpour
arXiv:2505.01361v3 Announce Type: replace
Abstract: Temporal difference (TD) learning is a foundational algorithm in reinforcement learning (RL). For nearly forty years, TD learning has served as a w...
By Hwanwoo Kim, Panos Toulis, Eric Laber
arXiv:2602. 02304v3 Announce Type: replace Abstract: Large-scale foundation models exhibit behavioral shifts when subjected to interventions such as scaling, fine-tuning, reinforcement learning with human feedback, or in-context learning.
By Martino Ciaperoni, Marzio Di Vece, Roberto Pellungrini, Luca Pappalardo, Fosca Giannotti, Francesco Giannini
arXiv:2608.23939v1 Announce Type: new
Abstract: Offline reinforcement learning is intrinsically multi-objective: a policy must remain compatible with the behavioral support of a fixed dataset while p...
By Xiewei Ni, Ruofeng Mei, Xiangyu Xu