…at inference time it receives only video frames: no target caption, no prompt, and no text encoder. Given a budget of k frames, PEEK predicts per-frame relevance scores and returns the…
…under strong angular intervention, a modest expansion of this capacity relieves competition between the steered concept and other context-relevant features, accounting for the observed stability gain. Paper: https://arxiv.org/abs…
…reasoning and then selects question-relevant frames adaptively, resulting in sub-quadratic attention cost with little to no accuracy loss. We arrange frames on a K{times}K grid and run lightweight…
…Siyuan Guo , , , , Abstract Deployment-time learning enables large language model agents to adapt continuously during operation through episodic memory and contextual bandit optimization, improving performance across diverse tasks. AI-generated summary Large…
…Hassan Ismkhan , Abstract RL4IL enables robust robotic manipulation under sensor dropout by using reinforcement learning to retrieve relevant demonstrations and cross-attention fusion to impute missing modalities without retraining. Generated by Qwen…
…More broadly, ALE is intended not merely as another leaderboard, but as an instrument for closing the gap between benchmark success and GDP-relevant impact. View arXiv page View PDF Project page…
…HExA iteratively designs and refines query-relevant experiments , learns a reusable library of composable skills from experience, and integrates experimental evidence to answer queries or take actions. HExA is training-free, compatible…
…Does the overhead of managing the routing and re-entry process add much to the compute time compared to just dropping the tokens entirely? I made a podcast on it with ResearchPod…
…Finally, we distill the causal base model into a few-step generator for real-time interactive generation . Extensive experiments demonstrate that CausalCine significantly outperforms autoregressive baselines and approaches the capability of bidirectional…
…top-ranked performance over GRPO and decoupled baselines, outperforms supervised alternatives, and remains competitive with strong zero-shot rerankers, with no architectural changes at inference time. View arXiv page View PDF Add…