KAIST
Language & Knowledge Lab

Following papers are accepted to ICML 2026:
- (Spotlight) Dynamics Reveals Structure: Challenging the Linear Propagation Assumption
- Early Decisions Matter: Proximity Bias and Initial Trajectory Shaping in Non-Autoregressive Diffusion Language Models
- Q-Flow: Stable and Expressive Reinforcement Learning with Flow-Based Policy
- Intrinsic Task Symmetry Drives Generalization in Algorithmic Tasks
- Decentralized Instruction Tuning: Conflict-Aware Splitting and Weight Merging

Following paper is accepted to ACL 2026:
- Can Large Language Models Keep Up? Benchmarking Online Adaptation to Continual Knowledge Streams

Following papers are accepted to ICLR 2026:
- The CoT Encyclopedia: Analyzing, Predicting, and Controlling how a Reasoning Model will Think
- Characterizing Pattern Matching and Its Limits on Compositional Task Structures
- TSLM: Tree-Structured Language Modeling for Divergent Thinking

Following paper is accepted to NeurIPS 2025:
- Reasoning Models Better Express Their Confidence

BiGGen Bench is awarded the Best Paper Award at NAACL 2025. [paper]