Research Feed Page 16

Rollplex increases GPU utilization during vision-language model post-training by overlapping prompt processing with rollout decoding to eliminate serial execution bottlenecks.

The paper introduces a method called SimpleOPD that distills advanced mathematical reasoning from a high-performance teacher model into smaller student models across different architectures.

The paper introduces a self-supervised method to improve vision model performance by distilling knowledge from a student to an EMA teacher without needing ground-truth labels or extra rewards.

The paper introduces a method to identify helpful reasoning in incorrect AI model messages that improves the final output of multi-agent systems.

Polaris is a multi agentic system designed to improve enterprise data querying and analysis through adaptive task coordination and specialized agent orchestration.

The paper introduces a toolkit that assesses robotic task execution by analyzing continuous progress curves rather than relying on binary success rates.

The authors introduce CPI-Bench, a comprehensive benchmark designed to evaluate image editing models across complex real-world scenarios, reasoning tasks, and multi-image workflows.

SheetCompass improves spreadsheet reasoning by representing data as a hierarchical graph rather than a flat string, and by coordinating multi-agent logic through a dual-level memory system.

The researchers developed an agent that creates and validates an executable code model at test time to solve rule-hidden games through planning and model repair.

The paper introduces a training-free framework that utilizes idle compute time between agent reasoning steps to perform parallel background tasks.

The paper identifies that Power Sampling causes accuracy drops by misaligning trajectory coverage, and introduces a Relative-Rank SoftSat method to improve consensus aggregation.

The paper introduces a Bayesian framework that optimizes LLM testing by dynamically stopping evaluations once statistical confidence thresholds are met.

DeaMoE is a decoding efficient architecture for Mixture of Experts models that reduces the amount of weight data loaded per inference step.

PACE-Bench provides a framework to evaluate how effectively AI agents can modify executable code to maintain functionality when physical environment parameters change.

Researchers built a new benchmark, RA-Bench, to systematically test how well current detection methods identify AI-generated videos during real-world social crises.

The paper introduces Intern-S2-Mobius, an architecture that decouples knowledge storage from reasoning to achieve faster inference speeds while maintaining high performance.

Wyvern is an agentic framework that automates the generation of multimodal reports by grounding text claims in web-sourced references and integrating relevant figures.

Marionette decouples game world state from visual appearance by using a two stage model to predict 3D geometry before synthesizing photorealistic video frames.

AgentRewind enables LLM agents to recover from errors during complex tasks by creating checkpoints and resuming from earlier states using past execution data.

CodeSIFT identifies batches of security-relevant, malicious code prompts by measuring how these inputs influence model behavior without requiring predefined vulnerability patterns.