Experience-Driven Dynamic Exits for LLMs with Reinforcement Learning
Mirrored from arXiv — NLP / Computation & Language for archival readability. Support the source by reading on the original site.
Computer Science > Computation and Language
Title:Experience-Driven Dynamic Exits for LLMs with Reinforcement Learning
Abstract:Large Language Models suffer from slow autoregressive inference. While self-speculative decoding accelerates this process, its efficiency is hampered by static configurations like fixed exit layers and speculation lengths. We reframe this optimization as a \textbf{Markov Decision Process} and propose \textbf{LEDE}, a framework that uses offline reinforcement learning. LEDE learns a policy to dynamically select the optimal exit layer and speculation length based on the local context of the generated sequence at each step, balancing computational cost and draft quality. Comprehensive evaluations on Llama-2 and Llama-3 models show LEDE achieves up to a $2.0\times$$\sim$$2.7\times$ speedup over autoregressive decoding and and provides an additional 17\% speedup over the static speculative baselines.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2606.03113 [cs.CL] |
| (or arXiv:2606.03113v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.03113
arXiv-issued DOI via DataCite (pending registration)
|
Access Paper:
- View PDF
- HTML (experimental)
- TeX Source
References & Citations
Bookmark
Bibliographic and Citation Tools
Code, Data and Media Associated with this Article
Demos
Recommenders and Search Tools
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
More from arXiv — NLP / Computation & Language
-
Hallucination Is Linearly Decodable from Mid-Layer Hidden States in Quantized LLMs
Jun 3
-
Filter, Then Reweight: Rethinking Optimization Granularity in On-Policy Distillation
Jun 3
-
IdiomX A Multilingual Benchmark for Idiom Understanding, Retrieval, and Interpretation
Jun 3
-
Greener Than Humans? Environmental Attitudes in Large Language Models
Jun 3
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.