# Student-Centered Distillation Narrows the Agentic Gap Between Small and Large LLMs

arXiv: https://arxiv.org/abs/2509.14257 · HTML: https://mlsanity.com/paper/2509.14257/

## Key facts

- Cited by 14 papers in the map (data through 14 Oct 2026).
- Placed in Reinforcement learning for LLM agents, one of 16 directions in LLM agents and tool use.
- Builds on 48 of its 59 references that are in the map.
- First posted on arXiv 12 Sep 2025.

## Placed in

- LLM agents and tool use / [Reinforcement learning for LLM agents](https://mlsanity.com/agents/reinforcement-learning-for-llm-agents/)

## Builds on (most cited first, 30 of 48)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Toolformer: Language Models Can Teach Themselves to Use Tools](https://mlsanity.com/paper/2302.04761/) (2023)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering](https://mlsanity.com/paper/1809.09600/) (2018)
- [FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning](https://mlsanity.com/paper/2307.08691/) (2023)
- [Emergent Abilities of Large Language Models](https://mlsanity.com/paper/2206.07682/) (2022)
- [HybridFlow: A Flexible and Efficient RLHF Framework](https://mlsanity.com/paper/2409.19256/) (2024)
- [LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://mlsanity.com/paper/2403.13372/) (2024)
- [The Rise and Potential of Large Language Model Based Agents: A Survey](https://mlsanity.com/paper/2309.07864/) (2023)
- [Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps](https://mlsanity.com/paper/2011.01060/) (2020)
- [AgentBench: Evaluating LLMs as Agents](https://mlsanity.com/paper/2308.03688/) (2023)
- [ALFWorld: Aligning Text and Embodied Environments for Interactive Learning](https://mlsanity.com/paper/2010.03768/) (2020)
- [Hindsight Experience Replay](https://mlsanity.com/paper/1707.01495/) (2017)
- [GAIA: a benchmark for General AI Assistants](https://mlsanity.com/paper/2311.12983/) (2023)
- [Measuring and Narrowing the Compositionality Gap in Language Models](https://mlsanity.com/paper/2210.03350/) (2022)
- [MuSiQue: Multihop Questions via Single-hop Question Composition](https://mlsanity.com/paper/2108.00573/) (2021)
- [Advantage-Weighted Regression: Simple and Scalable Off-Policy Reinforcement Learning](https://mlsanity.com/paper/1910.00177/) (2019)
- [Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm](https://mlsanity.com/paper/1712.01815/) (2017)
- [Executable Code Actions Elicit Better LLM Agents](https://mlsanity.com/paper/2402.01030/) (2024)
- [Behavioral Cloning from Observation](https://mlsanity.com/paper/1805.01954/) (2018)
- [Policy Distillation](https://mlsanity.com/paper/1511.06295/) (2015)
- [Search-o1: Agentic Search-Enhanced Large Reasoning Models](https://mlsanity.com/paper/2501.05366/) (2025)
- [The Lessons of Developing Process Reward Models in Mathematical Reasoning](https://mlsanity.com/paper/2501.07301/) (2025)

## Built on by (most cited first, 14 of 14)

- [A Survey of On-Policy Distillation for Large Language Models](https://mlsanity.com/paper/2604.00626/) (2026)
- [Multi-Rollout On-Policy Distillation via Peer Successes and Failures](https://mlsanity.com/paper/2605.12652/) (2026)
- [One-Way Policy Optimization for Self-Evolving LLMs](https://mlsanity.com/paper/2605.22156/) (2026)
- [From Scoring to Acting: Outcome-Verified Comparative Self-Distillation for LLM Agents](https://mlsanity.com/paper/2607.27937/) (2026)
- [When Agents Look the Same: Quantifying Distillation-Induced Similarity in Tool-Use Behaviors](https://mlsanity.com/paper/2604.21255/) (2026)
- [Learn from Weaknesses: Automated Domain Specialization for Small Computer-Use Agents](https://mlsanity.com/paper/2605.28775/) (2026)
- [It Takes 8 Tokens: Weak-to-Strong Off-Policy RL via Auxiliary Branches](https://mlsanity.com/paper/2607.16205/) (2026)
- [Signal Reshaping for GRPO in Weak-Feedback Agentic Code Repair](https://mlsanity.com/paper/2605.07276/) (2026)
- [Persistent Teacher Anchoring for Tool-Using Agents](https://mlsanity.com/paper/2609.04773/) (2026)
- [MInTRL: Off-policy Intervention can boost On-policy RL](https://mlsanity.com/paper/2609.12419/) (2026)
- [Can Terminal Agents Trust Their Own Verification? Diagnosing and Improving Self-Verification](https://mlsanity.com/paper/2609.38812/) (2026)
- [OverAct: Measuring and Mitigating Proactive Over-Authorization in LLM Tool-Calling Agents](https://mlsanity.com/paper/2610.01508/) (2026)
- [Beyond Corrected Actions: Training LLM Agents on Textual Critiques](https://mlsanity.com/paper/2610.19367/) (2026)
- [ActionOPD: Distilling Decisions, Not Tokens](https://mlsanity.com/paper/2610.19820/) (2026)

## Abstract (arXiv)

Large Language Model agents achieve strong performance on multi-step reasoning and tool-use tasks, but their impressive capabilities typically rely on extremely large backbones. Existing distillation approaches train smaller students to imitate full teacher trajectories, yet reasoning and knowledge gaps between the teacher and student can cause compounding errors. We propose SCoRe, a student-centered framework in which the student generates training trajectories and the teacher corrects only the earliest error, producing training data matched to the student's abilities and exposing specific weaknesses. The student is first fine-tuned on corrected trajectories. Subsequently, short-horizon reinforcement learning starts from the verified prefix preceding the earliest error, with target rewards assigned at that step. This design enables the student to solve problems through unconstrained RL exploration rather than teacher imitation, while the short-horizon setup improves training stability. On 12 challenging benchmarks, a 7B-parameter student distilled with SCoRe closes the agentic performance gap with a 72B-parameter teacher.