A multi-agent-driven automated pipeline for continuously evolving multimodal benchmarks.

Paper on using sparse autoencoders to analyze internal activations of Neural Quantum States.

A system for scalable oversight of coding agents using traditional software engineering constraints.

A framework for guiding symbolic solvers with symbol-equivariant recurrent reasoning models.

Uncertainty-aware multi-agent collaboration framework for reliable software development.

Bounded-memory testbed for evaluating long-horizon LLM agents.

An evaluation framework for testing and analyzing LLM capabilities.

A benchmark for evaluating multimodal agentic capabilities through game development tasks.

Tool for capturing natural-language dependency evidence and modeling skill-package-service graphs.

Test-time memory decay and anti-loop regularization for discrete VLN

CLI tool to search coding agent history on local machines.

A family of autonomous driving datasets designed through a strategic gap-identification framework.

A log-structured filesystem for S3

Framework for learning gait-aware quadruped locomotion with Temporal Logic Specifications.

Unrolled Phase and Amplitude Decomposition Network for improved image deblurring.

An architecture for self-evolving agents using steering adapters and verifier-in-the-loop mechanisms to prevent regressions.

Graph-native reasoning models fine-tuned with GRPO for traceable scientific hypothesis generation.

Real-Time Source-Free Object Detection framework using YOLOv10.

A multi-task benchmark for remote sensing change detection, captioning, and QA.