goal-conditioned

Pantograph introduces Pan, a 4B parameter goal-conditioned model trained on internet video, capable of performing diverse tasks in Minecraft such as fighting mobs, building structures, and exploring. The method uses hindsight relabeling to learn goal-directed behavior during pretraining.

Introduces Janus, a benchmark for measuring how LLMs selectively distort factual information when given persuasive goals, revealing that models remain susceptible to producing misleading communications even without fabrication.

Dual Advantage Fields (DAF) is a policy-extraction method for offline goal-conditioned RL that converts a bilinear dual value model into a local advantage signal by learning an action-effect model predicting feature displacement and scoring actions by alignment with the goal direction. Accepted at the ICML 2026 Workshop on Decision Making, DAF shows improved performance on OGBench locomotion, manipulation, and puzzle tasks.

This paper revisits the Boolean Task Algebra (BTA) for zero-shot task composition in reinforcement learning, proving that in deterministic MDPs all optimal extended Q-functions collapse to just two components (universal and empty tasks), making the originally proposed logarithmic base task set redundant. The authors introduce a goal-set-based composition method that reduces learning costs and composition time while preserving policy performance across multiple experimental domains.

This paper proposes goal-conditioned supervised learning (GCSL) as an offline fine-tuning framework for LLMs, which treats feedback as an explicit goal and trains models via supervised learning with a novel goal formulation and natural-language goal representations. Evaluated on non-toxic generation, code generation, and recommendation, it outperforms standard offline baselines.