Engineering
Revisiting Local Context for Long-Horizon Streaming 3D Reconstruction
ABot-Recon is a streaming 3D reconstruction model that caches only the preceding 11 frames of features rather than maintaining extensive…
Engineering
Fast Weight Attention for Continual Learning
This paper introduces the Falcon family of fast-weight attention mechanisms for continual learning, compressing expanding context into fixed-size recurrent states…
Engineering
Act with Intent: Distilling Behavior Intent for Vision-Language-Action Models
Intention Distillation (INDI) trains vision-language-action models using a frozen teacher vision-language model to distill behavioral intent into the action decoder,…
Engineering
LayerRecall: A State-Conditioned Memory Router for Long-Horizon Consistency in Video Generation
LayerRecall is a memory-routing mechanism for video generation that selectively retrieves historical context and injects it into specific transformer layers…
Engineering
Locate Anything in Videos: Rethinking Efficient Generative Spatio-Temporal Video Grounding
Parallel Tube Decoding (PTD) is a generative approach to spatio-temporal video grounding that decomposes localization into a temporal block followed…
Engineering
Blind Men and the Elephant: Probing the Epistemic Myopia of LLMs under Long-Tail Divergent Knowledge
ElephantBench is a benchmark of 1,094 questions testing whether LLMs can recall multiple divergent accounts of long-tail facts, built via…
Engineering
Paint What You See: Benchmarking Dexterous Visual Tool Use in Multimodal Agents
EASEL is a benchmark evaluating multimodal agents’ ability to perform dexterous visual tool use through reference-guided visual reconstruction tasks, accompanied…
Engineering
StepGuard: Learning Step-Level Guardrails with Scalable Supervision and Safety-Utility Balancing
StepGuard is a guard model that audits LLM-agent trajectories and monitors tool actions before execution, paired with StepGen, an automatic…
Engineering
Ring Forcing: Towards Precise Long-Term Memory for Autoregressive Video Diffusion
Ring Forcing addresses long-term memory limitations in autoregressive video diffusion using a ring-structured training strategy combined with compression and timestep…
Engineering
Video Generative Models as Geometry Learner
GeoNeXt repurposes pretrained video generative models for geometry-estimation tasks such as depth and surface-normal prediction by formulating geometry estimation as…
Engineering
PonderPounce: A Pretrained MLLM as an Episode Context Engine for Robot Control
PonderPounce is a robot-control system using a pretrained multimodal LLM as episodic memory: “Ponder,” a reasoning-focused MLLM, accumulates observations and…
Engineering
Rubric-to-Code Credit Assignment for Reinforcement Learning
Rubric-to-Code Credit Assignment (RCCA) is a reinforcement-learning framework that converts functional feedback from rubrics into localized optimization signals for generated…