16 papers
StateM: Reaching 95.3% Raw Accuracy, or a $15 Frontier Run, on Terminal-Bench 2.1 via Harness Scaling
Ziheng Qin, Yaxin Lu, Zhangyang Atlas Wang +1
Long-horizon agents can fail even when their underlying models can solve the constituent steps. They may lose track of mutable state, fail to reactivate lessons from earlier execut…
HumanNOVA: Photorealistic, Universal and Rapid 3D Human Avatar Modeling from a Single Image
Hezhen Hu, Wangbo Zhao, Lanqing Guo +6
In this paper, we present HumanNOVA, a photorealistic, universal, and rapid model for generating 3D human avatars from a single RGB image. Achieving both photorealism and generaliz…
Not All Synthetic Data Is Yours to Learn From
Sina Alemohammad, Li Chen, Richard G. Baraniuk +1
Can a language model improve from plain text sampled from itself, with no prompts, no teacher, no verifier, and no reward model? Yes, but only when the synthetic corpus is compatib…
When Is Rank-1 Steering Cheap? Geometry, Granularity, and Budgeted Search
John T. Robertson, Jianing Zhu, Haris Vikalo +1
Activation steering offers a lightweight way to control LLMs without retraining, but its effectiveness varies sharply across concepts. Prior work often reads this variability as ev…
Position: Weight Space Should Be a First-Class Generative AI Modality
Zhangyang Wang, Peihao Wang, Kai Wang
Neural network checkpoints have quietly become a large-scale data resource: millions of trained weight vectors now exist, each encoding task-, domain-, and architecture-specific kn…
Good Agentic Friends Do Not Just Give Verbal Advice: They Can Update Your Weights
Wenrui Bao, Huan Wang, Jian Wang +3
Multi-agent LLM systems usually collaborate by exchanging natural-language messages. This interface is simple and interpretable, but it forces each sender's intermediate computatio…
Lean Refactor: Multi-Objective Controllable Proof Optimization via Agentic Strategy Search
Jialin Lu, Soonho Kong, Rodrigo Stehling +4
We present Lean Refactor, a plug-and-play retrieval-augmented agentic framework for multi-objective, controllable, and version-robust refactoring of Lean proofs. LLM-generated proo…
Correct Answers from Sound Reasoning: Verifiable Process Supervision for Language Models
Kyuyoung Kim, Kevin Wang, Yunfei Xie +7
Training language models to produce both correct answers and sound reasoning remains an open challenge. Reinforcement learning with verifiable rewards typically optimizes only fina…
ResearchGPT: Benchmarking and Training LLMs for End-to-End Computer Science Research Workflows
Penghao Wang, Yuhao Zhou, Mengxuan Wu +12
As large language models (LLMs) advance, the ultimate vision for their role in science is emerging: we could build an AI collaborator to effectively assist human beings throughout…
Data Efficient Any Transformer-to-Mamba Distillation via Attention Bridge
Penghao Wang, Yuhao Zhou, Mengxuan Wu +3
State-space models (SSMs) have emerged as promising alternatives to Transformers for sequence modeling. However, training competitive SSMs from scratch remains computationally inte…
RAPID^3: Tri-Level Reinforced Acceleration Policies for Diffusion Transformer
Wangbo Zhao, Yizeng Han, Zhiwei Tang +7
Diffusion Transformers (DiTs) excel at visual generation yet remain hampered by slow sampling. Existing training-free accelerators - step reduction, feature caching, and sparse att…
Drag-and-Drop LLMs: Zero-Shot Prompt-to-Weights
Zhiyuan Liang, Dongwen Tang, Yuhao Zhou +11
Modern Parameter-Efficient Fine-Tuning (PEFT) methods such as low-rank adaptation (LoRA) reduce the cost of customizing large language models (LLMs), yet still require a separate o…
DD-Ranking: Rethinking the Evaluation of Dataset Distillation
Zekai Li, Xinhao Zhong, Samir Khaki +49
In recent years, dataset distillation has provided a reliable solution for data compression, where models trained on the resulting smaller synthetic datasets achieve performance co…
Make Optimization Once and for All with Fine-grained Guidance
Mingjia Shi, Ruihan Lin, Xuxi Chen +8
Learning to Optimize (L2O) enhances optimization efficiency with integrated neural networks. L2O paradigms achieve great outcomes, e.g., refitting optimizer, generating unseen solu…
Enhance-A-Video: Better Generated Video for Free
Yang Luo, Xuanlei Zhao, Mengzhao Chen +5
DiT-based video generation has achieved remarkable results, but research into enhancing existing models remains relatively unexplored. In this work, we introduce a training-free ap…
A Stitch in Time Saves Nine: Small VLM is a Precise Guidance for Accelerating Large VLMs
Wangbo Zhao, Yizeng Han, Jiasheng Tang +5
Vision-language models (VLMs) have shown remarkable success across various multi-modal tasks, yet large VLMs encounter significant efficiency challenges due to processing numerous…