Researchers present MaP-WAM, a framework for non-Markovian robotic manipulation that decomposes memory-dependent world-action modeling into memory-grounded planning and plan-conditioned execution.…
Researchers present X-AuT, a technique for making speech-focused large language models more efficient by compressing their audio encoders through progressive…
Researchers introduce Mi-Ripple, a restoration technique that targets degradation artifacts that accumulate from repeated cycles of AI-based image editing. The…
Researchers present SpatialBlock, an approach for improving how large vision-language models understand 3D spatial relationships from 2D images. The method…
The paper introduces Feedback-Enriched Environments (FEEs), which shift training focus from agent-side improvements to environment-side adaptations rather than relying solely…
The paper introduces Procedural Graphs, a framework that organizes procedural knowledge as structured (procedure, relation, procedure) triplets to guide LLM…
Researchers introduce NeoHorse-1, an approach to recursive self-improvement built on agentic post-training over a heterogeneous model pool with intelligent routing.…
Researchers present a system for accelerating reinforcement learning training by improving speculative decoding through online co-training of draft models. The…
The paper proposes BeaconKV, a training-free technique that compresses key-value caches during inference for large reasoning models. It identifies that…
Researchers present Gander, an end-to-end omni interaction agent that integrates multimodal perception, real-time conversation, and complex task execution within a…
Researchers analyzed 7.5 million single-model invocations and roughly 300,000 onchain actions across two deployed LLM trading-agent fleets over six months.…
Google DeepMind released AlphaGenome Atlas, a deep learning system that provides precomputed predictions for all 9 billion possible single-nucleotide variants…
