Energy-Structured Low-Rank Adaptation for Continual Learning.

Agentic model routing framework for coding tasks to optimize performance and cost.

Training framework for GUI Action Critic models to improve agent reliability.

Mechanistic evidence of attention head specialization under LLM jailbreak attacks.

Progressive multimodal search agent for Knowledge-Based Visual Question Answering (KB-VQA).

Code for Drop-Then-Recovery analysis protocol to measure redundancy in VLA models.

An open-source security monitoring platform (SIEM/XDR).

Pocket-sized open-source media server using ESP32-S3 to serve a browser-based library offline.

A full-stack React framework that focuses on type-safe routing and data loading.

A dual-channel model for combinatorial algorithm selection in the Maximum Clique Problem.

A deterministic control plane for managing LLM coding agents as a managed supply chain.

Human-in-the-loop agentic system for collaborative scientific visualization.

A retrieval memory system that maintains temporal validity to eliminate stale-fact errors for AI agents.

Research map for ML turning papers into markdown summaries and concept elements.

A framework for distilling and compiling agent traces into reusable procedural skills.

Transformer-based model for long-term time series forecasting using patch-mean decoupling.

Dynamic Preference Optimization framework for recommendation systems to prevent optimization collapse.

A tactile-language framework and dataset for scaling tactile commonsense reasoning in open-world settings.

A local debugger for visualizing and analyzing AI agent traces.

Open-source self-hosted platform for file storage, synchronization, and collaboration.