A self-evolving critic framework for calibrated step-level confidence estimation in LLM agents.

A codebook-oblivious vector index enabling deterministic per-tenant retrieval cost calculation.

A simulator for smart-greenhouse climate control and reward-component auditing.

Longitudinal multi-view breast cancer risk prediction model.

Code and results for evaluating text summarization abstractiveness metrics.

An AI voice agent for Asterisk/FreePBX that can be self-hosted for privacy and control.

A framework for stabilizing on-policy LLM RL with off-policy anchor samples

A diffusion-based optical flow framework using confidence-guided soft inpainting.

Research on training LLMs to reason in Japanese using GRPO, exploring performance vs English baselines.

An audio generation baseline demonstrating strong trade-offs between reference alignment and diversity in SFX morphing.

Benchmark and method for fine-grained text-to-video retrieval of camera-trap data.

An automatic rollout curation tool for segmenting robot trajectories to improve post-training data utilization.

An entity-oriented multi-media memory system for open-ended video streams.

Implementation of research on optimizing the placement of strong correctors in multi-agent swarms.

Task-conditioned synthetic data generation for agricultural ML prediction.

A live TUI switchboard for every signal on your Linux box

Agentic pipeline that reconstructs verifiable Milestone DAGs from commit logs for AI evaluation.

Project page for Multimodal Mamba for efficient text-to-video generation.

A benchmark for visual-centric reasoning on diverse map documents for LVLMs.

Agentic LLM system for automating warehouse operations using reasoning and task execution.