10 papers
LoopREG: Representation Supervision Enables Test-Time Recurrent Depth in Diffusion Transformers
Zipeng Wang, Anping Xiong, Jing Zhang +17
Reusing Transformer blocks offers a way to increase diffusion-model computation at test time, but naive recurrence can degrade generation beyond its training depth. We introduce Lo…
Is Multimodal Speculative Decoding Ready for Diffusion-Based Parallel Drafting? A Survey and Empirical Diagnosis
Yantao Li, Huanlin Gao, Fang Zhao +12
Speculative decoding accelerates autoregressive generation by allowing a lightweight drafter to propose future tokens while a target model verifies them in parallel. Its lossless g…
Scalable Question-Centric Text-to-Image Evaluation: Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing
Shaoan Zhao, Fang Zhao, Xueqiang Guo +9
Modern text-to-image (T2I) models often have similar total scores but different strengths, making practical selection difficult. Fine-grained benchmarks decompose prompts into ques…
Token-Operations-Oriented Inference Optimization Techniques for Large Models
Shiguo Lian, Kai Wang, Zhaoxiang Liu +23
Large model inference optimization serves as a key foundation for supporting the scalable, low-cost, and highly stable operation of large model services. Centered on token-oriented…
OTCache: Optimal Transport for Geometry-Aware Caching in Diffusion Models
Huanlin Gao, Fang Zhao, Qiang Hui +8
We propose OTCache, a training-free framework for accelerating diffusion sampling via caching schedule prediction. Existing graph-based caching methods reduce redundant computation…
How Far Can Unsupervised RLVR Scale LLM Training?
Bingxiang He, Yuxin Zuo, Zeyuan Liu +18
Unsupervised reinforcement learning with verifiable rewards (URLVR) offers a pathway to scale LLM training beyond the supervision bottleneck by deriving rewards without ground trut…
PaLMR: Towards Faithful Visual Reasoning via Multimodal Process Alignment
Yantao Li, Qiang Hui, Chenyang Yan +8
Reinforcement learning has recently improved the reasoning ability of Large Language Models and Multimodal LLMs, yet prevailing reward designs emphasise final-answer correctness an…
MeanCache: From Instantaneous to Average Velocity for Accelerating Flow Matching Inference
Huanlin Gao, Ping Chen, Fuyuan Shi +11
We present MeanCache, a training-free caching framework for efficient Flow Matching inference. Existing caching methods reduce redundant computation but typically rely on instantan…
LeMiCa: Lexicographic Minimax Path Caching for Efficient Diffusion-Based Video Generation
Huanlin Gao, Ping Chen, Fuyuan Shi +5
We present LeMiCa, a training-free and efficient acceleration framework for diffusion-based video generation. While existing caching strategies primarily focus on reducing local he…
HiMo-CLIP: Modeling Semantic Hierarchy and Monotonicity in Vision-Language Alignment
Ruijia Wu, Ping Chen, Fei Shen +8
Contrastive vision-language models like CLIP have achieved impressive results in image-text retrieval by aligning image and text representations in a shared embedding space. Howeve…