ByteDance Seed · 45 records

dateidtitleauthorsvenue↑
2026-11-202611.24809On-Policy Morphing into Hybrid Attention ModelsDisen Lan et al.arXiv12
2026-11-092610.41711Hierarchical Residuals: A Minimalist Decoupling of Residual Width from Compute WidthZiteng Wang et al.arXiv8
2026-11-052611.08089SAE Interprets MoE: Sparse AutoEncoders Reveal That MoE Experts Exhibit Human-Interpretable Fine-Grained SpecializationAnyuan Zhuo et al.arXiv7
2026-10-292610.40362Your LR Schedule May Not Be Principled: Effective Learning Rate Scheduling in Weight-Normalized TrainingQifan Yu et al.arXiv7
2026-10-232610.31305Cola2: Co-Adaptive Hierarchical Language Modeling with Continuous Semantic Compression and Balanced Discrete GenerationHongcan Guo et al.arXiv21
2026-10-152610.20399Cortex: Consolidating Context into Parametric MemoryJintian Zhang et al.arXiv70
2026-10-122610.12102Few-Step Generation via Data-Space IterationShanchuan Lin et al.arXiv3
2026-10-012609.30840Aligning One-Step Generative Models with Reward-Weighted Transport DistillationAustin Wang et al.arXiv2
2026-09-302609.36322Periodic Weak Spots: Phase Sensitivity from Chunked KV-Cache CompressionXingyu Zhu et al.arXiv36
2026-09-282609.31093Block Sparse Attention with Log-Linear ComplexityBohao Tang et al.arXiv32
2026-09-222609.22870Towards Full Pipeline FP8 Reinforcement Learning for LLMsFanchao Chen et al.arXiv14
2026-09-212609.20816Paint-Anything: Unified Any-Color Control for Image Generation and EditingJi Xie et al.arXiv50
2026-09-032609.01437HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?Yuhao Wu et al.arXiv88
2026-09-032608.31111Aspire: Can Models Self-Evolve from Vague Goals?Yuhao Wu et al.arXiv49
2026-09-032608.31100S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?Jiajun Shi et al.arXiv37
2026-09-022609.01343SMELT: Scaling Laws for Compute-Matched MoE Looped TransformersShaowen Wang et al.arXiv54
2026-09-012608.29335GenFirst: Generation Before Reconstruction for Stable End-to-End Latent Generative ModelingGuangting Zheng et al.arXiv72
2026-09-012608.30821Lucida: Parse, Generate, and Place for Composable Real-to-Sim Scene ModelingMinghan Qin et al.arXiv64
2026-08-262608.24646On-Policy Self-Distillation in Diffusion ModelsWei Zhou et al.arXiv68
2026-08-192608.17800StartupBench: Benchmarking General-Purpose Agents on Market-Validated End-to-End WorkflowsLiya Zhu et al.arXiv14
2026-08-172608.14071Scaling Domain Data Repetition in LLM PretrainingJingwei Li et al.arXiv16
2026-08-102608.07110Modular TTT: Rethinking Test-Time Training as Composable ModulesBohao Tang et al.arXiv11
2026-08-072608.05747GST-Bench: Can VLMs Develop Global Spatial Awareness from Video?Qifeng Zhang et al.arXiv25
2026-08-032607.29679Scaling Properties of Text Conditioning in Visual GenerationZilong Chen et al.arXiv30
2026-07-212607.18227FlowMimic: Mask-free Visual Editing and Generation with Pixel-pair Warped Flow Field for Online Video Editing Data Generation and Modality MimicryDingyun Zhang et al.arXiv52
2026-07-152607.11886Read It Back: Pretrained MLLMs Are Zero-Shot Reward Models for Text-to-Image GenerationRunhui Huang et al.arXiv57
2026-07-102607.08763OpenCoF: Learning to Reason Through Video GenerationXinyan Chen et al.arXiv24
2026-07-102607.06987UP: Unbounded Positive Asymmetric Optimization for Breaking the Exploration-Stability DilemmaChongyu Fan et al.arXiv8
2026-07-072607.05155EdgeBench: Unveiling Scaling Laws of Learning from Real-World EnvironmentsDeyao Zhu et al.arXiv15
2026-07-032606.30562Morphing into Hybrid Attention ModelsDisen Lan et al.arXiv49
2026-07-022607.00248Seed2.0 Model Card: Towards Intelligence Frontier for Real-World ComplexityBytedance SeedarXiv30
2026-06-292606.28133Translation as a Bridging Action: Transferring Manipulation Skills from Humans to RobotsSijin Chen et al.arXiv39
2026-06-262606.27377DanceOPD: On-Policy Generative Field DistillationWei Zhou et al.arXiv82
2026-06-252606.25331Improved Large Language Diffusion ModelsShen Nie et al.arXiv43
2026-06-242606.24742World Value Models for Robotic ManipulationZhihao Wang et al.arXiv8
2026-06-102606.11042Workflow-GYM: Towards Long-Horizon Evaluation of Computer-use Agentic tasks in Real-World Professional FieldsLiya Zhu et al.arXiv24
2026-06-102606.10650Dynamic Linear AttentionXin Wang et al.arXiv9
2026-06-012605.31604Representation Forcing for Bottleneck-Free Unified Multimodal ModelsYuqing Wang et al.arXiv61
2026-06-012605.31075Task-Focused Memorization for Multimodal AgentsTao Zou et al.arXiv39
2026-05-272605.26895Negligible in Size, Significant in Effect: On Scale Vectors in Large Language ModelsMingze Wang et al.arXiv17
2026-05-142605.13831Training Long-Context Vision-Language Models Effectively with Generalization Beyond 128K ContextZhaowei Wang et al.arXiv89
2026-05-082605.06548Continuous Latent Diffusion Language ModelHongcan Guo et al.arXiv89
2026-05-062605.02134Video Generation with Predictive LatentsYian Zhao et al.arXiv22
2026-05-042605.00503End-to-End Autoregressive Image Generation with 1D Semantic TokenizerWenda Chu et al.arXiv10
2026-05-012604.27505Leveraging Verifier-Based Reinforcement Learning in Image EditingHanzhong Guo et al.arXiv25