# Learning Like Humans: Advancing LLM Reasoning Capabilities via Adaptive Difficulty Curriculum Learning and Expert-Guided Self-Reformulation

Source: [Learning Like Humans: Advancing LLM Reasoning Capabilities via Adaptive Difficulty Curriculum Learning and Expert-Guided Self-Reformulation](<https://www.sota2.com/research/paper/68fa7b681e209131cea883cb>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T05:39:30.468Z

Authors: Enci Zhang, Xingang Yan, Wei Lin, Tianxiang Zhang, Qianchun Lu

Published: 2025-05-13T09:10:48.000Z

## Abstract

Despite impressive progress in areas like mathematical reasoning, large language models still face significant challenges in consistently solving complex problems. Drawing inspiration from key human learning strategies, we propose two novel strategies to enhance the capability of large language models to solve these complex problems. First, Adaptive Difficulty Curriculum Learning (ADCL) is a novel curriculum learning strategy that tackles the Difficulty Shift phenomenon (i.e., a model's perception of problem difficulty dynamically changes during training) by periodically re-estimating difficulty within upcoming data batches to maintain alignment with the model's evolving capabilities. Second, Expert-Guided Self-Reformulation (EGSR) is a novel reinforcement learning strategy that bridges the gap between imitation learning and pure exploration by guiding models to reformulate expert solutions within their own conceptual framework, rather than relying on direct imitation, fostering deeper understanding and knowledge assimilation. Extensive experiments on challenging mathematical reasoning benchmarks, using Qwen2.5-7B as the base model, demonstrate that these human-inspired strategies synergistically and significantly enhance performance. Notably, their combined application improves performance over the standard Zero-RL baseline by 10% on the AIME24 benchmark and 16.6% on AIME25.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Mathematical Reasoning on DAPO-17k](<https://www.sota2.com/research/sota/mathematical-reasoning-on-dapo-17k.md>): Mathematical Reasoning; DAPO-17K; AIME 2024 Score: 16.7.

- [Code Generation on CodeContests+](<https://www.sota2.com/research/sota/code-generation-on-codecontests-plus.md>): Code Generation; CodeContests+; LCBv6 Score: 37.2.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2505.08364>)

[PDF](<https://arxiv.org/pdf/2505.08364.pdf>)
