The paper introduces the reflexivity matrix Ω to capture how one agent’s deployment influences another’s best response in multi‑agent performative prediction. It identifies a spectral threshold—when the spectral radius of Ω exceeds one—separating stable learning from worst‑case learning obstruction. Below this threshold, repeated retraining converges geometrically to a unique stable equilibrium with bounded excess loss; above it, cumulative excess loss can grow linearly with the horizon.
By Ahmed Mahrous, Roberto Di Pietro
The paper investigates collusive behavior in a two‑player Almgren‑Chriss liquidation game using independent proximal policy optimisation agents. By training agents that learn a punitive mechanism, the study shows that when one agent deviates from the mean liquidation schedule, the other responds by accelerating liquidation, thereby negating the deviator’s gains while keeping its own payoff stable. The authors formalize checks confirming that punishment outweighs deviation gains and that behavioral changes are sufficient to offset losses, providing both behavioral and economic evidence of collusion.
By Christos Spyridon Koulouris, Carlo Campajola
arXiv:2610. 11475v1 Announce Type: new Abstract: Population gradient flow is a common tool for reasoning about how neural networks adapt, including after pretraining.
By Ruoyu Zhao, Mingxuan Zhang, Jianbo Dai, Jiaqi Wu, Chenyu Zhu, Tong Che
arXiv:2606. 03238v1 Announce Type: cross Abstract: Reinforcement learning from human feedback (RLHF) makes large-scale post-training possible by replacing an underspecified human objective with learned and scalable proxies.
By Zelalem Abahana
The paper investigates why reinforcement learning with self‑generated labels—specifically majority‑vote rewards—can recover most of the performance gains of training with verifiable rewards. It argues that the key factor is batch‑level updates: a prompt’s own gradient contributes less than 1% to its output change, so only coherent label errors across prompts influence learning. Experiments on five models show that majority‑vote updates align almost perfectly with gold updates, with misdirection disappearing as batch size grows and incoherent errors behaving like random noise.
By Dylan P. Wilson, Chengsong Huang, Jiaxin Huang
The paper proposes CANOPY, a minimalist reinforcement learning protocol that addresses two common pitfalls—signal starvation and policy drift—in outcome‑only RL for long‑horizon interactive tasks. By scaling same‑task exploration, keeping updates on‑policy, and anchoring updates with KL divergence, CANOPY enables a Qwen3‑14B agent to achieve top leaderboard results on the AppWorld coding benchmark without auxiliary supervision or elaborate scaffolding. The approach also improves performance on SWE‑bench for a Qwen3.5‑9B model.
By Liming Pu, Xiaoxia Li, Yifu Liu, Teng Cao, Bin Yang
The paper demonstrates that safety mechanisms for autonomous large language model agents fail to compose across iterative loops, as trajectory‑scoped monitors cannot detect attacks whose evidence is spread over multiple iterations. It introduces LoopHarness, a system that maintains a persistent, non‑decaying safety state across loops, bounding unauthorized actions with a constant that does not grow with the number of iterations. The authors provide a comprehensive evaluation protocol, including attacks that require cross‑iteration evidence, module ablations, and adaptive white‑box red‑team testing.
By Chenhao Wu, Haoxuan Jia, Yang Liu, Yingguang Yang, Yuhan Lin, Chongyang Zhang, Hao Zheng, Yulin Huang, Jianshen Zhang, Yongzhi Qi, Shang Luo, Kefu Xu, Jifeng Zhu, Bin Chong
Population gradient flow is a common tool for reasoning about how neural networks adapt, including after pretraining. We show that it can mispredict finite-batch stochastic gradient descent (SGD) qualitatively, and we trace the discrepancy to a specific mechanism.
arXiv:2601. 17454v2 Announce Type: replace-cross Abstract: Centralized value learning underlies a broad class of multi-agent reinforcement learning methods, but its claimed advantage is typically evaluated in settings that confound coordination structure with function approximation and partial observability.
By Muhammad Ahmed Atif, Nehal Naeem Haji, Mohammad Shahid Shaikh, Muhammad Ebad Atif
The paper studies a stationary decentralized Markov game where a focal agent experiences drifting rewards and dynamics due to learning peers, framing this as an agent‑centric continual reinforcement‑learning problem. It introduces the concept of an invariant core—maximal abstract patterns common to many successful trajectories—and proves a worst‑case conditioning theorem linking trajectory‑law drift to success coverage. The authors provide theoretical guarantees for survival horizon, first‑exit law, and regret, and validate their predictions with solvable models and empirical studies in continual control, cue‑MNIST, and Level‑Based Foraging.
By Dane Malenfant
arXiv:2607. 26313v1 Announce Type: cross Abstract: Agentic systems act, so a defect in the evidence they retrieve becomes a wrong action with a currency cost.
By Gaston Besanson
arXiv:2607. 17136v1 Announce Type: cross Abstract: Agentic computer-use RL is reported in single runs, and those numbers mislead.
By Barada Sahu (Cabal AI), Shivesh Pandey (Para AI)