Papers

Filtered to benchmarks · clear filter

Browse by term

continual learning 64reinforcement learning 34large language models 17benchmarking 12vision-language models 10generative models 8language models 8video generation 7multimodal models 6natural language processing 6robotics 6world models 6benchmarks 5diffusion models 5on-policy distillation 5policy optimization 5scalability 5self-distillation 5vision-language-action models 5autoregressive models 4computer vision 4diffusion transformers 4LLMs 4multimodal large language models 4verifiable rewards 4attention mechanisms 3embodied intelligence 3image editing 3long-term memory 3multimodal learning 3

Matching papers

Pass the Baton: Trajectory-Relayed On-Policy Distillation

28 upvotes · 28 JUL 2026 · Haolei Xu, Xiaowen Xu, Haiwen Hong et al.

This paper addresses a problem in on-policy distillation where a student model can get stuck on a wrong path, and proposes a new method called Relay-OPD that helps the student model recover by briefly taking over at certain points to produce a new trajectory. Practitioners might care about this because it could lead to better performance and more efficient training in models like language generators or math solvers.

on-policy distillationtrajectory-based methodscontinual learningreinforcement learningbenchmarks

HPD-Parsing: Hierarchical Parallel Document Parsing

8 upvotes · 21 JUL 2026 · Shu Wei, Jingjing Wu, Lingshu Zhang et al.

This paper introduces HPD-Parsing, a new approach to document parsing that uses hierarchical parallel decoding to improve efficiency and throughput. Practitioners in natural language processing and computer vision might care because it could lead to faster and more accurate document parsing models.

vision-language modelsdocument parsinghierarchical parallel decodingautoregressive modelsbenchmarks

Reasoning Denoiser: Denoising Reasoning Traces for Hallucination Detection in Large Reasoning Models

6 upvotes · 24 JUL 2026 · Junlin Fang, Do Nguyen-Thanh, Xiaogang Xu et al.

This paper develops a method to clean up noisy signals in reasoning traces of large models to improve detection of hallucinations, which are false answers produced by models. Practitioners might care about this because it could lead to more accurate models that produce reliable answers.

continual learningreasoning modelshallucination detectiondenoisingnatural language processingbenchmarks

Codifying the Judge: Scalable Evaluation via Program Distillation

6 upvotes · 29 MAY 2026 · Tzu-Heng Huang, Shengqi Qiu, Frederic Sala

This paper proposes a way to make automated evaluation systems more efficient, transparent, and reliable by distilling the decision logic of large language models into smaller, programmatic judges that can be easily inspected and edited. Practitioners might care because this approach could help reduce costs and improve the scalability of automated evaluation systems.

program distillationLLM-as-a-judgescalable evaluationtransparent decision-makingcontinual learningbenchmarks