Automatic Curriculum Expert Iteration for Reliable LLM Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhao, Zirui, Dong, Hanze, Saha, Amrita, Xiong, Caiming, Sahoo, Doyen
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913746553667584
author Zhao, Zirui
Dong, Hanze
Saha, Amrita
Xiong, Caiming
Sahoo, Doyen
author_facet Zhao, Zirui
Dong, Hanze
Saha, Amrita
Xiong, Caiming
Sahoo, Doyen
contents Hallucinations (i.e., generating plausible but inaccurate content) and laziness (i.e. excessive refusals or defaulting to "I don't know") persist as major challenges in LLM reasoning. Current efforts to reduce hallucinations primarily focus on factual errors in knowledge-grounded tasks, often neglecting hallucinations related to faulty reasoning. Meanwhile, some approaches render LLMs overly conservative, limiting their problem-solving capabilities. To mitigate hallucination and laziness in reasoning tasks, we propose Automatic Curriculum Expert Iteration (Auto-CEI) to enhance LLM reasoning and align responses to the model's capabilities--assertively answering within its limits and declining when tasks exceed them. In our method, Expert Iteration explores the reasoning trajectories near the LLM policy, guiding incorrect paths back on track to reduce compounding errors and improve robustness; it also promotes appropriate "I don't know" responses after sufficient reasoning attempts. The curriculum automatically adjusts rewards, incentivizing extended reasoning before acknowledging incapability, thereby pushing the limits of LLM reasoning and aligning its behaviour with these limits. We compare Auto-CEI with various SOTA baselines across logical reasoning, mathematics, and planning tasks, where Auto-CEI achieves superior alignment by effectively balancing assertiveness and conservativeness. The code is available at https://github.com/SalesforceAIResearch/Auto-CEI .
format Preprint
id arxiv_https___arxiv_org_abs_2410_07627
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Automatic Curriculum Expert Iteration for Reliable LLM Reasoning
Zhao, Zirui
Dong, Hanze
Saha, Amrita
Xiong, Caiming
Sahoo, Doyen
Machine Learning
Artificial Intelligence
Computation and Language
Hallucinations (i.e., generating plausible but inaccurate content) and laziness (i.e. excessive refusals or defaulting to "I don't know") persist as major challenges in LLM reasoning. Current efforts to reduce hallucinations primarily focus on factual errors in knowledge-grounded tasks, often neglecting hallucinations related to faulty reasoning. Meanwhile, some approaches render LLMs overly conservative, limiting their problem-solving capabilities. To mitigate hallucination and laziness in reasoning tasks, we propose Automatic Curriculum Expert Iteration (Auto-CEI) to enhance LLM reasoning and align responses to the model's capabilities--assertively answering within its limits and declining when tasks exceed them. In our method, Expert Iteration explores the reasoning trajectories near the LLM policy, guiding incorrect paths back on track to reduce compounding errors and improve robustness; it also promotes appropriate "I don't know" responses after sufficient reasoning attempts. The curriculum automatically adjusts rewards, incentivizing extended reasoning before acknowledging incapability, thereby pushing the limits of LLM reasoning and aligning its behaviour with these limits. We compare Auto-CEI with various SOTA baselines across logical reasoning, mathematics, and planning tasks, where Auto-CEI achieves superior alignment by effectively balancing assertiveness and conservativeness. The code is available at https://github.com/SalesforceAIResearch/Auto-CEI .
title Automatic Curriculum Expert Iteration for Reliable LLM Reasoning
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2410.07627