One-Pass to Reason: Token Duplication and Block-Sparse Mask for Efficient Fine-Tuning on Multi-Turn Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Goru, Ritesh, Mehta, Shanay, Jain, Prateek
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912476862349312
author Goru, Ritesh
Mehta, Shanay
Jain, Prateek
author_facet Goru, Ritesh
Mehta, Shanay
Jain, Prateek
contents Fine-tuning Large Language Models (LLMs) on multi-turn reasoning datasets requires N (number of turns) separate forward passes per conversation due to reasoning token visibility constraints, as reasoning tokens for a turn are discarded in subsequent turns. We propose duplicating response tokens along with a custom attention mask to enable single-pass processing of entire conversations. We prove our method produces identical losses to the N-pass approach while reducing time complexity from $O\bigl(N^{3}\bigl)$ to $O\bigl(N^{2}\bigl)$ and maintaining the same memory complexity for a transformer based model. Our approach achieves significant training speedup while preserving accuracy. Our implementation is available online (https://github.com/devrev/One-Pass-to-Reason).
format Preprint
id arxiv_https___arxiv_org_abs_2504_18246
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle One-Pass to Reason: Token Duplication and Block-Sparse Mask for Efficient Fine-Tuning on Multi-Turn Reasoning
Goru, Ritesh
Mehta, Shanay
Jain, Prateek
Computation and Language
Artificial Intelligence
Machine Learning
Fine-tuning Large Language Models (LLMs) on multi-turn reasoning datasets requires N (number of turns) separate forward passes per conversation due to reasoning token visibility constraints, as reasoning tokens for a turn are discarded in subsequent turns. We propose duplicating response tokens along with a custom attention mask to enable single-pass processing of entire conversations. We prove our method produces identical losses to the N-pass approach while reducing time complexity from $O\bigl(N^{3}\bigl)$ to $O\bigl(N^{2}\bigl)$ and maintaining the same memory complexity for a transformer based model. Our approach achieves significant training speedup while preserving accuracy. Our implementation is available online (https://github.com/devrev/One-Pass-to-Reason).
title One-Pass to Reason: Token Duplication and Block-Sparse Mask for Efficient Fine-Tuning on Multi-Turn Reasoning
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2504.18246