Saved in:
Bibliographic Details
Main Authors: Chanda, Prateek, Sureka, Saral, Chatterjee, Parth Pratim, Killamsetty, Krishnateja, Nayak, Nikhil Shivakumar, Ramakrishnan, Ganesh
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.12612
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913978573127680
author Chanda, Prateek
Sureka, Saral
Chatterjee, Parth Pratim
Killamsetty, Krishnateja
Nayak, Nikhil Shivakumar
Ramakrishnan, Ganesh
author_facet Chanda, Prateek
Sureka, Saral
Chatterjee, Parth Pratim
Killamsetty, Krishnateja
Nayak, Nikhil Shivakumar
Ramakrishnan, Ganesh
contents The performance of finetuned large language models (LLMs) hinges critically on the composition of the training mixture. However, selecting an optimal blend of task datasets remains a largely manual, heuristic driven process, with practitioners often relying on uniform or size based sampling strategies. We introduce TASKPGM, a principled and scalable framework for mixture optimization that selects continuous task proportions by minimizing an energy function over a Markov Random Field (MRF). Task relationships are modeled using behavioral divergences such as Jensen Shannon Divergence and Pointwise Mutual Information computed from the predictive distributions of single task finetuned models. Our method yields a closed form solution under simplex constraints and provably balances representativeness and diversity among tasks. We provide theoretical guarantees, including weak submodularity for budgeted variants, and demonstrate consistent empirical improvements on Llama 2 and Mistral across evaluation suites such as MMLU and BIGBench. Beyond performance, TASKPGM offers interpretable insights into task influence and mixture composition, making it a powerful tool for efficient and robust LLM finetuning.
format Preprint
id arxiv_https___arxiv_org_abs_2507_12612
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learning What Matters: Probabilistic Task Selection via Mutual Information for Model Finetuning
Chanda, Prateek
Sureka, Saral
Chatterjee, Parth Pratim
Killamsetty, Krishnateja
Nayak, Nikhil Shivakumar
Ramakrishnan, Ganesh
Machine Learning
Artificial Intelligence
68T50
I.2.7; I.2.6; I.2.4
The performance of finetuned large language models (LLMs) hinges critically on the composition of the training mixture. However, selecting an optimal blend of task datasets remains a largely manual, heuristic driven process, with practitioners often relying on uniform or size based sampling strategies. We introduce TASKPGM, a principled and scalable framework for mixture optimization that selects continuous task proportions by minimizing an energy function over a Markov Random Field (MRF). Task relationships are modeled using behavioral divergences such as Jensen Shannon Divergence and Pointwise Mutual Information computed from the predictive distributions of single task finetuned models. Our method yields a closed form solution under simplex constraints and provably balances representativeness and diversity among tasks. We provide theoretical guarantees, including weak submodularity for budgeted variants, and demonstrate consistent empirical improvements on Llama 2 and Mistral across evaluation suites such as MMLU and BIGBench. Beyond performance, TASKPGM offers interpretable insights into task influence and mixture composition, making it a powerful tool for efficient and robust LLM finetuning.
title Learning What Matters: Probabilistic Task Selection via Mutual Information for Model Finetuning
topic Machine Learning
Artificial Intelligence
68T50
I.2.7; I.2.6; I.2.4
url https://arxiv.org/abs/2507.12612