Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

ConMeZO: Adaptive Descent-Direction Sampling for Gradient-Free Finetuning of Large Language Models

About

Zeroth-order or derivative-free optimization (MeZO) is an attractive strategy for finetuning large language models (LLMs) because it eliminates the memory overhead of backpropagation. However, it converges slowly due to the inherent curse of dimensionality when searching for descent directions in the high-dimensional parameter space of billion-scale LLMs. We propose ConMeZO, a novel zeroth-order optimizer that accelerates convergence by adaptive directional sampling. Instead of drawing the direction uniformly at random, ConMeZO restricts the sampling to a cone centered around a momentum estimate. This concentrates the search in directions where the true gradient is more likely to lie and thus reduces the effect of high dimensions. We prove that ConMeZO achieves the same worst-case convergence rate as MeZO. Empirically, when finetuning LLMs on natural language tasks, ConMeZO is up to 2X faster than MeZO while retaining the low-memory footprint of zeroth-order methods.

Lejs Deen Behric, Liang Zhang, Bingcong Li, Kiran Koshy Thekumparampil• 2025

Related benchmarks

TaskDatasetResultRank
Natural Language InferenceSNLI (test)
Accuracy81.9
694
Natural Language InferenceRTE
Accuracy76.2
590
Question AnsweringBoolQ--
317
Question ClassificationTREC
Accuracy90.4
262
Word Sense DisambiguationWiC
Avg Accuracy58.31
261
Natural Language InferenceSNLI
Accuracy82
196
Sentiment AnalysisSST-5 (test)
Accuracy48.9
177
Question ClassificationTREC (test)
Accuracy90
128
Sentiment AnalysisSST-5
Accuracy50
123
Reading ComprehensionDROP
F1 Score26.53
96
Showing 10 of 18 rows

Other info

Follow for update