arXiv:2607. 27422v1 Announce Type: new Abstract: Good action rankings do not make a contrastive critic safe to maximize.
arXiv:2602. 05459v2 Announce Type: replace Abstract: Offline goal-conditioned reinforcement learning (GCRL) is typically benchmarked by the best tuned success rate of each method.
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.
The paper introduces GeZo‑SAC, a variant of Soft Actor‑Critic that incorporates geometric representations (zonotopes) to adapt critic pessimism based on state and action. Each critic predicts a zonotope whose width is subtracted from the critic value as a pessimistic offset, and the disagreement between critics is measured and used to blend the critics via a log‑sum‑exp aggregation. Experiments on MuJoCo‑v5 locomotion tasks show GeZo‑SAC achieving the highest mean return on Ant‑v5 and Hopper‑v5, while also reducing actuator work and action effort and maintaining near‑zero overestimation frequency.
arXiv:2608. 01205v1 Announce Type: new Abstract: Recent offline reinforcement learning methods increasingly rely on expressive generative policies and specialized value-guidance mechanisms.
ARC‑Bench is a new benchmark that tests whether frozen JEPA‑style latent world models can correctly rank candidate actions by latent distance. The study finds that the assumption of latent rankability fails dramatically in both navigation and manipulation tasks, with the top‑scored actions often being suboptimal. Closed‑loop replanning masks this defect, but reducing replanning frequency reveals the underlying ranking failures.
arXiv:2609.00892v1 Announce Type: new Abstract: Rubric-based reinforcement learning decomposes open-ended instructions into prompt-specific, flexible rubrics, making it better suited than reinforceme...
arXiv:2608. 08491v1 Announce Type: new Abstract: Reward models are a bottleneck for reinforcement learning in embodied AI.
arXiv:2606. 11891v1 Announce Type: cross Abstract: Multi-objective reinforcement learning for humanoid robots must coordinate locomotion and manipulation within a single policy.
arXiv:2608. 00832v1 Announce Type: new Abstract: Structured plan-generation agents are often evaluated as if a plan has quality in isolation, yet many realistic planning tasks require asking how a candidate behaves when another agent can search for responses.
CARE‑VI introduces a framework for improving value targets in off‑policy actor‑critic learning by combining Conservative Adaptive Ranking and Screening (CARS), Selector‑Evaluator Value Assessment (SEVA), and Dynamic Adaptive Risk‑aware Enhancement (DARE). CARS limits candidate actions to a budgeted prefix and expands it only when uncertainty exceeds a threshold; SEVA orders candidates with selector critics and reviews their values with an evaluator critic, capping the value at the selector reference; DARE adjusts residual corrections based on candidate reliability and signal gaps. Theoretical analysis bounds errors in each component, and empirical tests on SAC, TD3, and TD7 across four MuJoCo tasks show CARE‑VI consistently outperforms baselines in mean return.
The paper presents a constrained command‑conditioned Proximal Policy Optimization (PPO) executor for the MicroRTS real‑time strategy environment, enabling a single execution policy to follow diverse strategic commands. A Thompson‑sampling bandit strategist selects command tuples based on in‑game observations of the opponent’s strategy, rather than relying on opponent identity. In controlled experiments, the strategist‑executor system outperforms a flat PPO baseline against three of the four strongest opponents on a training map, including two held‑out opponents, while showing no significant difference against the remaining opponents.
arXiv:2608. 02181v1 Announce Type: new Abstract: Proximal Policy Optimization (PPO) for large language models typically trains its critic by mean-squared-error (MSE) regression on scalar value targets.