A plug-and-play alignment correction method for enhancing LLM safety without fine-tuning.

AI-facing clinical record substrate using append-only Pods and SQLite projections for deterministic grounding.

A director-orchestrated multi-stage agent for chest X-ray interpretation using tool coordination and collaborative reasoning.

Robot foundation models with extended visuomotor context for long-horizon tasks.

Interactive website showing cost-success evaluations of offensive and defensive security agents.

A comprehensive benchmark for evaluating general AI agents across diverse application scenarios.

Open-source quantum software platform for quantum machine learning.

A code-controlled agent loop for evaluating and improving LLM candidate generation through structured feedback.

Code for training small LLMs to introspect their internal activations.

A fully on-device graph-based RAG framework for mobile personal assistants.

A framework for accelerating Diffusion MLLMs via MLP sparsity-aware truncation.

Framework for contextualized evaluations of MLLMs through dynamic, multi-round interactions.

Open-source infrastructure for AI agent evaluation using multi-juror, consensus-based scoring.

An MCP-based interface exposing power system simulation capabilities to AI agents.

Implementation of Dialogue Place Recognition for interactive geo-localization.

Encoder-free ECG-Language Models (ELF) for automated ECG interpretation.

A framework for discovering ordinary differential equations with LLM-based evaluation.

Algorithm-hardware co-design for accelerating Video Diffusion Transformers via latent space correlations.

An explainable gate that blocks LLM prompt injection

Open-source web UI for power grid modeling and power flow.