# The Realignment Problem: When Right becomes Wrong in LLMs

arXiv: https://arxiv.org/abs/2511.02623 · HTML: https://mlsanity.com/paper/2511.02623/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in LLM preference optimization methods, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 30 of its 37 references that are in the map.
- First posted on arXiv 4 Nov 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, 30 of 30)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [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)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Machine Unlearning](https://mlsanity.com/paper/1912.03817/) (2019)
- [A General Language Assistant as a Laboratory for Alignment](https://mlsanity.com/paper/2112.00861/) (2021)
- [A General Theoretical Paradigm to Understand Learning from Human Preferences](https://mlsanity.com/paper/2310.12036/) (2023)
- [Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2307.15217/) (2023)
- [Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning](https://mlsanity.com/paper/2404.05868/) (2024)
- [Who's Harry Potter? Approximate Unlearning in LLMs](https://mlsanity.com/paper/2310.02238/) (2023)
- [Persona Vectors: Monitoring and Controlling Character Traits in Language Models](https://mlsanity.com/paper/2507.21509/) (2025)
- [Large Language Model Unlearning](https://mlsanity.com/paper/2310.10683/) (2023)
- [PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference](https://mlsanity.com/paper/2406.15513/) (2024)
- [Influence Functions in Deep Learning Are Fragile](https://mlsanity.com/paper/2006.14651/) (2020)
- [In-Context Unlearning: Language Models as Few Shot Unlearners](https://mlsanity.com/paper/2310.07579/) (2023)
- [Provably Robust DPO: Aligning Language Models with Noisy Feedback](https://mlsanity.com/paper/2403.00409/) (2024)
- [WorldValuesBench: A Large-Scale Benchmark Dataset for Multi-Cultural Value Awareness of Language Models](https://mlsanity.com/paper/2404.16308/) (2024)
- [DailyDilemmas: Revealing Value Preferences of LLMs with Quandaries of Daily Life](https://mlsanity.com/paper/2410.02683/) (2024)
- [GenARM: Reward Guided Generation with Autoregressive Reward Model for Test-time Alignment](https://mlsanity.com/paper/2410.08193/) (2024)
- [Inverse Constitutional AI: Compressing Preferences into Principles](https://mlsanity.com/paper/2406.06560/) (2024)
- [Linear Representations of Political Perspective Emerge in Large Language Models](https://mlsanity.com/paper/2503.02080/) (2025)
- [ValueBench: Towards Comprehensively Evaluating Value Orientations and Understanding of Large Language Models](https://mlsanity.com/paper/2406.04214/) (2024)
- [UnStar: Unlearning with Self-Taught Anti-Sample Reasoning for LLMs](https://mlsanity.com/paper/2410.17050/) (2024)
- [Multi-Reference Preference Optimization for Large Language Models](https://mlsanity.com/paper/2405.16388/) (2024)
- [Correcting Large Language Model Behavior via Influence Function](https://mlsanity.com/paper/2412.16451/) (2024)
- [Leveraging Machine Unlearning for Cost-Efficient Preference Alignment](https://mlsanity.com/paper/2504.06659/) (2025)

## Abstract (arXiv)

Post-training alignment of large language models (LLMs) relies on large-scale human annotations guided by policy specifications that change over time. Cultural shifts, value reinterpretations, and regulatory or industrial updates make static alignment increasingly brittle. As policies evolve, deployed models can diverge from current alignment objectives, creating an Alignment-Reality Gap that is difficult to audit or correct. Existing remediation typically requires re-annotation under revised guidelines, which introduces systematic challenges, including guideline ambiguity, annotator interpretation drift, and reduced consistency at scale. We introduce TRACE (Triage and Re-align by Alignment Conflict Evaluation), a framework that transforms realignment into a structured optimization problem over existing data without requiring fresh human annotation. Leveraging a stronger model as a proxy judge, TRACE operates via a three-stage pipeline: (1) triaging preference pairs into inversion, suppression, or retention categories based on alignment conflicts; (2) computing an alignment impact score via bi-level optimization to prioritize high-leverage samples; and (3) executing updates using a hybrid objective that combines relational losses (e.g., IPO) for preference inversion and punitive losses (e.g., NPO) for response suppression. Experiments on Qwen2.5-7B, Gemma-2-9B, and Llama-3.1-8B demonstrate robust realignment on synthetic benchmarks and the PKU-SafeRLHF dataset without degrading general utility. This work provides a scalable approach for LLM realignment under evolving data annotation policies and alignment guidelines. We release our code: https://respailab.github.io/TRACE/