Flexible point cloud scene generation framework using Bird's-Eye-View (BEV) prior and optimal transport.

Self-hosted long-term memory backend for AI.

Two-stage unlearning framework for scalable privacy in recommender systems.

A calibration benchmark evaluating verbalized confidence estimates in 15 frontier LLMs.

Open-source macOS video editor built for AI.

A manually curated dataset and probe for assessing contextual entrainment in VLMs.

Memory-based Personalized Recommendation Tool learning via autonomous language Agents.

A creative plot generation framework utilizing SLMs and preference-based alignment.

Research examining KANs for small language models and their interpretability vs MLP performance.

A spatio-temporal forecasting architecture using Kolmogorov-Arnold Network modules.

A benchmark for evaluating LLM procedural reasoning in expert investment philosophy.

A fast-slow dual-system architecture for long-horizon aerial vision-language navigation.

Bibliometric analysis and interactive atlas of structural concentration in AI bias research.

A fast algorithm for extracting multi-dimensional refusal subspaces from LLM activations.

Complementary memory systems for test-time adaptation of vision-language models.

Vocabulary-Aligned Sparse Autoencoder for naming SAE dictionary directions.

Open source cycling trainer analytics for FTP and power curve tracking.

A typed hallucination auditing and multi-agent debate framework for legal AI.

Efficient test-time Q-steering for flow policies to improve RL action generation.

A semi-supervised framework for scaling LLM reasoning using a reasoning-correctness classifier.