Soft Reasoning: Navigating Solution Spaces in Large Language Models through Controlled Embedding Exploration

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhu, Qinglin, Zhao, Runcong, Yan, Hanqi, He, Yulan, Chen, Yudong, Gui, Lin
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909785319800832
author Zhu, Qinglin
Zhao, Runcong
Yan, Hanqi
He, Yulan
Chen, Yudong
Gui, Lin
author_facet Zhu, Qinglin
Zhao, Runcong
Yan, Hanqi
He, Yulan
Chen, Yudong
Gui, Lin
contents Large Language Models (LLMs) struggle with complex reasoning due to limited diversity and inefficient search. We propose Soft Reasoning, an embedding-based search framework that optimises the embedding of the first token to guide generation. It combines (1) embedding perturbation for controlled exploration and (2) Bayesian optimisation to refine embeddings via a verifier-guided objective, balancing exploration and exploitation. This approach improves reasoning accuracy and coherence while avoiding reliance on heuristic search. Experiments demonstrate superior correctness with minimal computation, making it a scalable, model-agnostic solution. The code is released at https://github.com/alickzhu/Soft-Reasoning.
format Preprint
id arxiv_https___arxiv_org_abs_2505_24688
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Soft Reasoning: Navigating Solution Spaces in Large Language Models through Controlled Embedding Exploration
Zhu, Qinglin
Zhao, Runcong
Yan, Hanqi
He, Yulan
Chen, Yudong
Gui, Lin
Computation and Language
Large Language Models (LLMs) struggle with complex reasoning due to limited diversity and inefficient search. We propose Soft Reasoning, an embedding-based search framework that optimises the embedding of the first token to guide generation. It combines (1) embedding perturbation for controlled exploration and (2) Bayesian optimisation to refine embeddings via a verifier-guided objective, balancing exploration and exploitation. This approach improves reasoning accuracy and coherence while avoiding reliance on heuristic search. Experiments demonstrate superior correctness with minimal computation, making it a scalable, model-agnostic solution. The code is released at https://github.com/alickzhu/Soft-Reasoning.
title Soft Reasoning: Navigating Solution Spaces in Large Language Models through Controlled Embedding Exploration
topic Computation and Language
url https://arxiv.org/abs/2505.24688