SoftmaxGRPO: Learning to Reason using Softmax Advantage Group Estimation

Confererence on Language Modeling, COLM 2026

HypoExplore: Agentic Discovery with Active Hypothesis Exploration

Findings of the Conference on Empirical Methods in Natural Language Processing. EMNLP 2026.

Learning to Reason by Analogy via Retrieval-Augmented Reinforcement Fine-Tuning

Findings of the Conference on Empirical Methods in Natural Language Processing. EMNLP 2026.

Beyond Referring Expressions: Scenario Comprehension Visual Grounding

Conference on Empirical Methods in Natural Language Processing. EMNLP 2026.

One Model, Many Budgets: Elastic Latent Interfaces for Diffusion Transformers

IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2026.

ProxyThinker: Test-Time Guidance through Small Visual Reasoners

International Conference on Learning Representations. ICLR 2026.

MetaEmbed: Scaling Multimodal Retrieval at Test-Time with Flexible Late Interaction

International Conference on Learning Representations. ICLR 2026.

SportR: A Benchmark for Multimodal Large Language Model Reasoning in Sports

International Conference on Learning Representations. ICLR 2026.