ByteDance-Seed ORG

Open-source unified multimodal model

Depth Anything 3

VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo

Seed1.5-VL, a vision-language foundation model designed to advance general-purpose multimodal understanding and reasoning, achieving state-of-the-art performance on 38 out of 60 public benchmarks.

Distributed Compiler based on Triton for Parallel Systems

No description.

Repo for SeedVR2 (ICLR2026) & SeedVR (CVPR2025 Highlight)

No description.

No description.

[CVPR 2025] VideoWorld is a simple generative model that learns purely from unlabeled videos—much like how babies learn by observing their environment.

Seed-Coder is a family of lightweight open-source code LLMs comprising base, instruct and reasoning models, developed by ByteDance Seed.

[ICLR 2026] Trace Anything: Representing Any Video in 4D via Trajectory Fields

No description.

EvaLearn is a pioneering benchmark designed to evaluate large language models (LLMs) on their learning capability and efficiency in challenging tasks.

[ICLR 2026] DAComp: Benchmarking Data Agents across the Full Data Intelligence Lifecycle

[ICML 2025 Spotlight] ShadowKV: KV Cache in Shadows for High-Throughput Long-Context LLM Inference

ByteCheckpoint: An Unified Checkpointing Library for LFMs

The codebase of Cola DLM

No description.

No description.

AHN: Artificial Hippocampus Networks for Efficient Long-Context Modeling

Resources for our paper: "Agent-R: Training Language Model Agents to Reflect via Iterative Self-Training"

No description.

Code for paper: [ICLR2025 Oral] FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference

Sim-to-real and CDM inference code for ManipAsInSim project.

A quantization algorithm for LLM

No description.

WideSearch: Benchmarking Agentic Broad Info-Seeking

Official implementation of Chain-of-Action: Trajectory Autoregressive Modeling for Robotic Manipulation. Accepted in NeurIPS 2025.

Implementation for "The Scalability of Simplicity: Empirical Analysis of Vision-Language Learning with a Single Transformer"

Stable-DiffCoder is a family of lightweight open-source code DLLMs(diffusion large language models) comprising base and instruct models, developed by ByteDance Seed.

No description.

No description.

Official code for VINCIE: Unlocking In-context Image Editing from Video

No description.

No description.

CVPR 2026 (Highlight); Spatial Intelligence; MLLMs

[Arxiv 2025] ByteMorph: Benchmarking Instruction-Guided Image Editing with Non-Rigid Motions

No description.

official implementation of paper SDP4Bit: Toward 4-bit Communication Quantization in Sharded Data Parallelism for LLM Training

No description.

[ICCV 2025] Deeply Supervised Flow-Based Generative Models

JIT-compiled GPU kernels for quantum chemistry

Generative foundation model for cryo-EM density maps.

Official implementation of SimFlow

[Neurips 2025] Official Implementation of ConfRover models

No description.

Joint Selection for Large-Scale Pre-Training Data via Policy Gradient-based Mask Learning

No description.

This is a minimal artifact with state-of-the-art speculative decoding as described in the SwiftSpec paper [ASPLOS' 26].

No description.

No description.

The codes for paper "It Takes Two to Entangle" (ASPLOS'26)

No description.

No description.

[SIGGRAPH 2026] SIMART: Decomposing Monolithic Meshes into Sim-ready Articulated Assets via MLLM

FELIS: Free Energy of Ligand-protein InteractionS

No description.

Development repository for the Triton language and compiler