HuggingFace

UK AISI and EvalEval are collaborating to openly share AI evaluation results using a standardized schema and platform, enhancing reproducibility and transparency in benchmarking for AI models.

Hugging Face's transformers library now supports GGUF models from llama.cpp, enabling efficient local inference on consumer hardware through familiar APIs.

Jun Kim, creator of oMLX, joins Hugging Face to support the MLX community, enhancing stability and development for local AI on Apple Silicon.

The paper presents a physics-inspired approach to pruning LLM blocks by modeling block removal as a constrained binary optimization problem mapped to an Ising glass, achieving significant compression gains without benchmarking each configuration.

The paper presents ACLArena, a framework for evaluating Agent Continual Learning in multi-stage post-training, analyzing forgetting and generalization mechanisms, and proposing an improved ACL recipe using offline replay and LoRA experts.

This paper introduces Complex KDA, an enhanced version of Kimi Delta Attention that combines a delta-rule transformation with a reflection to achieve greater expressivity, outperforming Transformers in some tasks while maintaining efficiency, with open-source code and models available.

EdgeGen is a synthetic task generation framework that creates database-grounded edge-case tasks to improve tool-calling agents through fine-tuning and harness optimization, demonstrating consistent performance improvements.

This paper introduces SVEET, a framework for high-quality streaming video editing that leverages a pretrained video diffusion model to enable auto-regressive editing with real-time performance on a single GPU.

The paper introduces THAW-VLA, a method that distills world-model representations into Vision-Language-Action models for robotics, enhancing robustness and performance on simulation and real hardware without additional inference overhead.

The paper presents VideoGen-Agent, a reinforcement learning-based multimodal agent that coordinates tools for video generation, significantly improving performance on the new VABench benchmark.

Jev-Mem introduces an agentic memory architecture inspired by System-One/System-Two cognition, enhancing efficiency and effectiveness for long-horizon AI agents with improved scores and faster operations.

The paper introduces GameHorizon Suite, a unified data and evaluation framework for assessing AI models' capabilities in gameplay across multiple temporal horizons, featuring an annotation pipeline, large-scale dataset, and reproducible benchmark.

This paper introduces Harness-Zero, a method for distilling optimized agent harnesses into large language models via agent-as-harness, significantly boosting task performance in knowledge work, tool use, and science domains even after specialized harnesses are removed.

onPanda is an interactive tool that uses token-level correction to efficiently annotate LLM alignment data and agent trajectories, reducing median annotation time by 52%.

The paper introduces an information-efficiency ratio (IER) for optimizing token selection in sparse on-policy distillation, demonstrating that using only 1% of tokens can achieve performance comparable to full supervision.

CARE is a framework for Vision-Language-Action policies that enhances robotic manipulation by learning from execution failures to generate corrective actions, demonstrating improved success rates in simulations and real-world tasks.

Document Retrieval-Aware Chunking (D-RAC) is a method that normalizes enterprise documents to PDF, converts them to retrieval-optimized Markdown using a multimodal LLM, and chunks them efficiently, significantly reducing token usage and costs compared to agentic chunking.

WorldCrafter is a video world model that learns a camera-queryable implicit 3D-aware memory for consistent and camera-controllable streaming scene exploration from a single image or text prompt.

This paper introduces Regularized Recursive Self-Improvement (RRSI) for AI agent harnesses, which applies regularization to prevent overfitting during recursive evolution, demonstrating performance gains on multiple benchmarks.

Hugging Face releases tokenizers v1, a major performance update for the tokenization library, with benchmarks showing significant speed improvements over previous versions.