 TagRouter: Learning Route to LLMs through Tags for Open-Domain Text Generation TasksAnnual Meeting of the Association for Computational Linguistics (ACL), 2025 |
Brevity is the soul of sustainability: Characterizing LLM response lengthsAnnual Meeting of the Association for Computational Linguistics (ACL), 2025 |
 Well Begun is Half Done: Low-resource Preference Alignment by Weak-to-Strong DecodingAnnual Meeting of the Association for Computational Linguistics (ACL), 2025 |
![Learning to Insert [PAUSE] Tokens for Better Reasoning](https://cdn.parameterlab.de/resized-images/aHR0cHM6Ly9jZG4ucGFyYW1ldGVybGFiLmRlL3BhcGVycy8yNTA2LjAzNjE2L3BhZ2VzL3BhZ2UuanBn?w=256&q=75&f=webp) Learning to Insert [PAUSE] Tokens for Better ReasoningAnnual Meeting of the Association for Computational Linguistics (ACL), 2025 |
 CLaSp: In-Context Layer Skip for Self-Speculative DecodingAnnual Meeting of the Association for Computational Linguistics (ACL), 2025 |
 Sparse-to-Dense: A Free Lunch for Lossless Acceleration of Video Understanding in LLMsAnnual Meeting of the Association for Computational Linguistics (ACL), 2025 |
 Inference Compute-Optimal Video Vision Language ModelsAnnual Meeting of the Association for Computational Linguistics (ACL), 2025 |
