Papers

Filtered to representation learning · 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

INTACT: Isomorphic Intent-to-Action Learning for Search-Free World Models

13 upvotes · 28 JUL 2026 · Junhan Sun, Hao Zhao, Guofeng Zhang

This paper introduces INTACT, a new method for training world models that can perform search-free actions without needing to test them. Practitioners might care because INTACT can improve the efficiency and effectiveness of world models in real-world applications.

continual learningisomorphic learningworld modelsaction modelslatent spacerepresentation learning

AutoIndex: Learning Representation Programs for Retrieval

9 upvotes · 21 JUL 2026 · Sam O'Nuallain, Nithya Rajkumar, Ramya Narayanasamy et al.

This paper introduces AutoIndex, a framework that learns to transform raw documents into representations for retrieval systems, allowing for more flexible and effective indexing. Practitioners may care about AutoIndex because it can improve the quality of search results in complex information retrieval tasks.

continual learningrepresentation learningretrieval systemsprogram synthesisvalidation-guided search

dRAE: Representation Autoencoder with Hyper-Spherical Codes

9 upvotes · 24 JUL 2026 · Tianren Ma, Lin Long, Chuyan Chen et al.

This paper develops a new method for quantizing high-dimensional visual representations into discrete codes that can be used with language models, allowing for more efficient and scalable training. Practitioners might care because it could enable the use of large language models on devices with limited memory.

representation learningquantizationautoencodershyper-spherical codesangular routing