A benchmark for evaluating long-horizon webpage generation from structural and functional perspectives.

A financial reasoning agent framework with selective experience memory for multimodal reasoning.

A framework for lightweight web data extraction using query-aware DOM pruning and scraper synthesis.

Code for sub-semantic image segmentation using DETECTURE to resolve language-to-mask failure modes.

A pipeline to turn real mathematical proofs into automatically checkable masked-step tasks.

Visual-native multimodal deep search agent using active visual reasoning.

Implementation of Fractured Sampling for efficient LLM reasoning trajectories.

A 3D point-cloud environment for probing and improving VLM view planning capabilities.

Framework for runtime model selection for Compound AI systems based on SLOs.

A lightweight implicit reasoning paradigm for LLM-based generative recommendation.

SOP enforcement mechanism tailored for open-source agentic browsers.

Reasoning tokens in superposition for faster LLM inference via token compression.

An open-source ListenBrainz-compatible scrobbler for tracking and visualizing music listening data.

Open-source coding model with a trillion-parameter MoE architecture focused on direct implementation.

An end-to-end robot learning stack for Vision-Language-Action models.

Forward-only LLM tuning using Local Credit Assignment to avoid repeated backpropagation.

An open-source framework to reduce semantic convergence and increase diversity in LLM deployments.

A retrieval-based framework for selecting visually relevant historical frames as dynamic sinks in long video generation.

LLM-driven system for translating natural language into validated MCMC samplers.

Code for equivariant linear layers based on irreducible representations and Schur's lemma.