aiqbench analysis
LEAF introduces a retrospective tree-based reinforcement learning method that recovers structural credit assignment in speech-aware models without incurring online branching costs. By grouping responses by shared prefixes and assigning span-level advantages, the approach enables smaller models to outperform full-parameter baselines on speech QA and translation tasks. This efficiency gain represents a significant advancement for resource-constrained deployment of specialized speech-language models.
LEAF: Growing Trees Without Branching for Speech-Aware Large Language Model Post-Training
arXiv:2606.07610v1 Announce Type: new Abstract: State-of-the-art GRPO-style methods for speech-aware large language model post-training suffer from coarse credit assignment, broadcasting the same terminal-reward advantage to every token in a response. This ignores useful structure within rollout batches, where speech-conditioned completions often share prefixes before diverging at important decisions. We propose Low-rank Exploration with Adaptive Forking (LEAF), a retrospective tree-based RL method that recovers this structure without online branching or additional decoding. LEAF samples complete responses, selects high-surprisal boundaries, groups responses by shared prefixes, and assigns span-level advantages using descendant rewards. We theoretically justify LEAF's span-level credit assignment and boundary-selection design. Empirically, LEAF improves over GRPO across speech question answering and speech translation benchmarks under the same rollout and low-rank adaptation budget. Notably, smaller LEAF-trained models outperform current state-of-the-art, full-parameter baselines.
Abstract:State-of-the-art GRPO-style methods for speech-aware large language model post-training suffer from coarse credit assignment, broadcasting the same terminal-reward advantage to every token in a response. This ignores useful structure within rollout batches, where speech-conditioned completions often share prefixes before diverging at important decisions. We propose Low-rank Exploration with Adaptive Forking (LEAF), a retrospective tree-based RL method that recovers this structure without online branching or additional decoding. LEAF samples complete responses, selects high-surprisal boundaries, groups responses by shared prefixes, and assigns span-level advantages using descendant rewards. We theoretically justify LEAF's span-level credit assignment and boundary-selection design. Empirically, LEAF improves over GRPO across speech question answering and speech translation benchmarks under the same rollout and low-rank adaptation budget. Notably, smaller LEAF-trained models outperform current state-of-the-art, full-parameter baselines.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
arXiv:2606.07610 [cs.LG]
(or arXiv:2606.07610v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2606.07610 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
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.
Full text under the publisher's open licence · 226 words
Related coverage
LEAF: Growing Trees Without Branching for Speech-Aware Large Language Model Post-Training
arXiv · Oct 7, 2026
PagedAttention vs. RadixAttention: Optimizing LLM KV Cache Management
Analytics Vidhya · Aug 21, 2026
SPIR-V on ROCm: A Portable IR for AMD GPUs
AMD · Jul 20, 2026
Notes on Algorithms
LessWrong · Jun 10, 2026