# Fine-Tuning LLMs with Fine-Grained Human Feedback on Text Spans

arXiv: https://arxiv.org/abs/2512.23693 · HTML: https://mlsanity.com/paper/2512.23693/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in LLM preference optimization methods, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 21 of its 28 references that are in the map.
- First posted on arXiv 29 Dec 2025.

## Placed in

- LLM alignment / RLHF / preference optimisation / [LLM preference optimization methods](https://mlsanity.com/alignment/llm-preference-optimization-methods/)

## Builds on (most cited first, 21 of 21)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [Large Language Models Can Self-Improve](https://mlsanity.com/paper/2210.11610/) (2022)
- [Multi-News: a Large-Scale Multi-Document Summarization Dataset and Abstractive Hierarchical Model](https://mlsanity.com/paper/1906.01749/) (2019)
- [UltraFeedback: Boosting Language Models with Scaled AI Feedback](https://mlsanity.com/paper/2310.01377/) (2023)
- [Fine-Grained Human Feedback Gives Better Rewards for Language Model Training](https://mlsanity.com/paper/2306.01693/) (2023)
- [Smaug: Fixing Failure Modes of Preference Optimisation with DPO-Positive](https://mlsanity.com/paper/2402.13228/) (2024)
- [HelpSteer: Multi-attribute Helpfulness Dataset for SteerLM](https://mlsanity.com/paper/2311.09528/) (2023)
- [Provably Robust DPO: Aligning Language Models with Noisy Feedback](https://mlsanity.com/paper/2403.00409/) (2024)
- [Anchored Preference Optimization and Contrastive Revisions: Addressing Underspecification in Alignment](https://mlsanity.com/paper/2408.06266/) (2024)
- [Improving Reward Models with Synthetic Critiques](https://mlsanity.com/paper/2405.20850/) (2024)
- [Self-Boosting Large Language Models with Synthetic Preference Data](https://mlsanity.com/paper/2410.06961/) (2024)
- [Curry-DPO: Enhancing Alignment using Curriculum Learning & Ranked Preferences](https://mlsanity.com/paper/2403.07230/) (2024)
- [Beyond Imitation: Leveraging Fine-grained Quality Signals for Alignment](https://mlsanity.com/paper/2311.04072/) (2023)
- [WildFeedback: Aligning LLMs With In-situ User Interactions And Feedback](https://mlsanity.com/paper/2408.15549/) (2024)
- [AI-Slop to AI-Polish? Aligning Language Models through Edit-Based Writing Rewards and Test-time Computation](https://mlsanity.com/paper/2504.07532/) (2025)
- [Finding the Sweet Spot: Preference Data Construction for Scaling Preference Optimization](https://mlsanity.com/paper/2502.16825/) (2025)
- [Data-Centric Human Preference with Rationales for Direct Preference Alignment](https://mlsanity.com/paper/2407.14477/) (2024)

## Built on by (most cited first, 1 of 1)

- [When Edit Localization Amplifies Relative Selection Bias: Gradient Geometry, Target Mismatch, and Importance Weighting](https://mlsanity.com/paper/2609.13709/) (2026)

## Abstract (arXiv)

We present a method and dataset for fine-tuning language models with preference supervision using feedback-driven improvement chains. Given a model response, an annotator provides fine-grained feedback by marking ``liked'' and ``disliked'' spans and specifying what they liked or disliked about them. The base model then rewrites the disliked spans accordingly, proceeding from left to right, forming a sequence of incremental improvements. We construct preference pairs for direct alignment from each adjacent step in the chain, enabling the model to learn from localized, targeted edits. We find that our approach outperforms direct alignment methods based on standard A/B preference ranking or full contrastive rewrites, demonstrating that structured, revision-based supervision leads to more efficient and effective preference tuning.