Research Feed Page 32

The researchers introduced a system called DEC that improves table parsing by decomposing, enhancing, and correcting errors through visual consistency checks.

ArchAgent v2 uses an evolutionary search process to automate the design of complex data prefetchers that exceed existing championship-level performance.

The SHE framework allows LLM agent safety systems to automatically evolve over time by analyzing failure trajectories to refine safety boundaries.

SymDiag improves LLM reasoning reliability by compiling natural language chains of thought into symbolic logic for automated diagnosis and repair.

RynnValue is an open-source foundation model that improves robotic task success rates by predicting the remaining distance to a goal using visual observations.

Motif 3 is a 314 billion parameter model that utilizes a specialized architecture and multi-stage post-training to improve intelligence and task generalization.

OasisKV uses predictive prefetching to move necessary KV cache data from high capacity memory to HBM, allowing for larger decode batches and longer context without hitting the memory wall.

Ouroboros introduces a self-developing agent architecture that treats code evolution as a task to adapt tools and core logic autonomously while maintaining strict safety standards.

The paper introduces Sci-VBench, a benchmark designed to test if generative video models can accurately simulate scientific mechanisms and causal dynamics.

OpenVisTool introduces a training method that teaches models to use external visual tools only when necessary, improving performance over fixed image encoding.

The paper introduces a controlled marketplace environment to evaluate the long term business performance and decision making capabilities of LLM agents.

The paper introduces Evo-Bench to evaluate how effectively large language models can autonomously evolve and improve the code harnesses that control agent behavior.

Researchers discovered an architectural flaw in how major LLM providers handle encrypted reasoning traces, allowing them to decrypt and expose proprietary data.

The researchers developed D-SCAN, a method to detect malicious documents injected into RAG pipelines by identifying specific anomalies in how the model distributes its attention.

The paper introduces SWE-Bench ProMax, a new, expert-curated benchmark designed to evaluate AI coding agents on complex, large-scale, multilingual refactoring tasks.

Macaron-V1 introduces a framework for deploying persistent agent models that update themselves through specialized adapters and recursive self-improvement loops.

The authors introduce GeoBenchLLM, a unified benchmark consisting of 421,041 questions across twelve datasets to evaluate LLM performance on geographic tasks.

The authors introduce PSG-JEPA, a model that adds physical grounding to latent world models to improve performance in real-world robotic manipulation tasks.

The paper introduces ArchEGraph, a dataset that maps building geometry and climate conditions to thermal energy loads to improve building energy modeling through graph-based machine learning.

The paper introduces the PHASE-Tree framework to manage character personality updates across long narrative sequences in role-playing agents.