The paper proposes a method for selectively querying language‑model advice in reinforcement learning by predicting the value of potential responses and only querying when the expected benefit outweighs the cost. It introduces a certified, response‑contingent metareasoning framework that guarantees near‑optimal advice usage under certain assumptions, and demonstrates that a calibrated controller with Qwen2.5 advisors can improve task performance while drastically reducing the number of advice calls on the BabyAI benchmark.
By Ibne Farabi Shihab, Md Najmus Swaqeeb, Abu Sa-Adat Mohamed Moon-Im Al Ahsan
The paper introduces PORA, a method that teaches large language model agents to decide when to allocate more reasoning steps versus fewer, based on the expected downstream benefit. PORA trains by running each state twice—once with short reasoning and once with long reasoning—and learns to predict the value of extended reasoning. During inference, a controller uses this prediction and remaining budget information to choose the appropriate reasoning length without running both options, improving performance across diverse tasks.
By Bo Yan, Qi Zhang, Zhen Tan, Tianlong Chen, Song Wang
arXiv:2608. 11560v1 Announce Type: new Abstract: Personalizing marketing messages with contextual multi-armed bandits (CMABs) drives real business value, yet the objective that ultimately matters - a downstream conversion - is observed only weeks later, too late to drive online learning.
By Sang Su Lee, Vineeth Loganathan, Shishir Dash, Vijay Raghavan
arXiv:2607. 26253v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) is bottlenecked by rollout generation, yet many sampled prompts produce saturated groups (all responses correct or all incorrect) whose zero reward variance yields no policy-gradient signal.
By Pixel Nomand, Elena Voss, Marcus Hale, Sofia Reyes
arXiv:2608. 05519v1 Announce Type: new Abstract: Agent benchmarks usually measure task completion and treat resource use as an auxiliary statistic.
By Jie Wu, Ming Gong, Feixiang Cheng, Qinqin Zhao
arXiv:2610.02828v1 Announce Type: new
Abstract: Adaptive LLM reinforcement-learning post-training changes multiple training actuators online, including rollout temperature, group size, clipping, KL r...
By Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Daren Zha, Jun Xiao
arXiv:2607. 13389v1 Announce Type: new Abstract: Reinforcement Learning (RL) post-training is increasingly used to adapt foundation models for reasoning, planning, and feedback-driven robot-learning pipelines, but constrained post-training resources are often summarized by a single total FLOP budget.
By Patrick Wilhelm, Odej Kao
Personalizing marketing messages with contextual multi-armed bandits (CMABs) drives real business value, yet the objective that ultimately matters - a downstream conversion - is observed only weeks later, too late to drive online learning. Teams therefore train the bandit on a fast proxy reward, and separately must judge whether a contextual bandit is worth its complexity over sending one best message.
arXiv:2609.37125v1 Announce Type: new
Abstract: Prospective memory allows an agent to retain an intention tied to a future condition, but the stored intention does not reveal whether that condition c...
By Zhengkun Di, Bin Shi, Kai Sun, Yiming Xu, Bo Dong
StoreBench is a live‑commerce environment where an autonomous agent runs a mid‑size online apparel store on a production‑grade backend, facing continuous customer orders, supplier price changes, and market shocks. The agent uses the same 29 merchant tools as a human operator, with a windowed operation budget that ties simulated time to actions, ensuring that model latency does not affect the simulation. Seven large language models were evaluated across 11 scenarios of 30–45 days and a full simulated year, but none matched the scripted smart‑triage policy, and human experts outperformed all models.
By Daksh Raghuvanshi, Ved Vedere, Yifan Wang
The paper investigates the problem of sharing a single critic across multiple parallel environments in reinforcement learning. It shows that when environments assign different expected returns to the same state, a shared critic must reconcile conflicting value targets, which can distort advantage estimates and misguide policy updates. The authors propose a simple fix—providing the critic with the environment index—demonstrating through bandit models and experiments on CartPole, MuJoCo, BipedalWalker, and 16 Procgen games that this conditional critic stabilizes learning and boosts returns, achieving a 40.8% improvement in aggregate normalized return on unseen levels.
By Zhenya Liu, Yang Meng, Zhuokai Zhao, Xuefeng Liu, Yuxin Chen
The paper introduces Discord, a method that uses reward disagreement to evaluate and compare harness edits via exact paired inference and early stopping, ensuring only genuinely differing outcomes influence decisions. Building on this, BRACE (Bank‑calibrated Rollout Allocation under Correlated Episodes) forecasts mixed‑reward groups by combining historical task evidence, uncertainty in success rates, and a correlated reward model, thereby allocating generation budgets more efficiently. Experiments on Qwen3‑VL, BFCL, and NESTFUL demonstrate that BRACE outperforms baseline methods with fewer new training rollouts and maintains strong performance without additional training.
By Hieu Tran, Zonghai Yao, Sunjae Kwon, Hong Yu