The paper introduces BALIGN, a balanced data selection strategy designed to reduce catastrophic forgetting—referred to as the alignment tax—in large language models during preference-based alignment. By analyzing preference optimization gradients, the authors identify three data-centric features that influence parameter drift: the reference model's log-probability margin, token length differences between chosen and rejected responses, and TF‑IDF similarity to general capability corpora. BALIGN aggregates these features into a composite risk score to filter out high-risk preference samples, thereby preserving foundational capabilities while maintaining alignment gains with minimal computational overhead.
By Minsu Kim, Jianxun Lian, Xing Xie, Steven Euijong Whang
Aligning large language models to human preferences is crucial for real-world deployment but frequently incurs an alignment tax, leading to the catastrophic forgetting of pre-trained general capabilit...
arXiv:2606. 01060v1 Announce Type: cross Abstract: Preference alignment has substantially improved the observable behavior of large language models, yet it remains unclear what alignment changes internally.
By Partha Pratim Saha, Samarth Raina, Mayur Parvatikar, Amit Dhanda, Vinija Jain, Aman Chadha, Amitava Das
The paper extends mechanistic interpretability of large language models by modeling concepts as low‑dimensional non‑linear manifolds rather than linear subspaces. It introduces a concept‑based alignment (CBA) score to compare these manifolds across layers and models, revealing block structures in intermediate layers, a shift from syntax‑dominated to mixed syntactic‑semantic concepts, and training‑dependent multilingual sharing. The study also shows that alignment patterns differ across model families and training stages, with adjacent stages aligning more closely than distant ones.
By Tido Specht, Elias Benedict Krey, Nils Neukirch, Nils Strodthoff
The paper introduces DSPA, a dynamic sparse autoencoder (SAE) steering technique that aligns language model outputs with user preferences during inference, avoiding costly weight updates. DSPA constructs a conditional-difference map from preference triples to adjust token-active latents, improving MT‑Bench scores and matching AlpacaEval performance on models like Gemma‑2 and Qwen3 while preserving accuracy. It demonstrates robustness with limited preference data, outperforms the two‑stage RAHF‑SCIT pipeline in FLOPs, and reveals that preference directions are largely driven by discourse and stylistic cues.
By James Wedgwood, Aashiq Muhamed, Mona T. Diab, Virginia Smith
arXiv:2609.05899v1 Announce Type: cross
Abstract: Aligning large language models with human preferences remains a challenge, primarily due to the critical role of preference data quality in effective...
By Peng Lai, He Zhu, Zhiwen Ruan, Dongdong Zhang, Yun Chen, Peng Li, Furu Wei, Yang Liu, Guanhua Chen
arXiv:2608.23149v1 Announce Type: cross
Abstract: The alignment of Large Language Models heavily relies on English-centric high-quality preference data, which often leads to suboptimal performance in...
By Seungyoon Lee, Minhyuk Kim, Jungseob Lee, Heuiseok Lim
arXiv:2609.06893v1 Announce Type: cross
Abstract: Direct preference optimization DPO is a promising offline approach for aligning large language models (LLMs) due to its simplicity, computational eff...
By Wenbo Zhang, Wenzhuo Zhou, Hengrui Cai, Zhengling Qi
The paper introduces GAP-DPO, a method for personalizing large language models by selecting preference pairs based on gradient alignment with user utility. It formalizes personalized preference learning as a geometry‑aligned optimization problem, showing that off‑policy sampling can shift DPO updates from error correction to reinforcement when preference margins align with utility gradients. Experiments demonstrate that GAP‑DPO improves stylistic fidelity, preference alignment, and overall generation quality over standard DPO variants.
By Ruoming Jin, Xinyu Li, Hao Zhou, Jianfeng Zhu, Ruixin Guo, Feodor Dragan, Lei Xu, Haixun Wang, Yang Zhou
arXiv:2504. 06659v2 Announce Type: replace-cross Abstract: Despite advances in Preference Alignment (PA) for Large Language Models (LLMs), mainstream methods like reinforcement learning with human feedback face notable challenges.
By Xiaohua Feng, Yuyuan Li, Huwei Ji, Jiaming Zhang, Li Zhang, Tianyu Du, Chaochao Chen
arXiv:2608. 19598v1 Announce Type: cross Abstract: Direct Preference Optimization (DPO) has emerged as an effective approach for aligning large language models (LLMs) with human preferences.
By Jiawei Feng, Jiancan Wu, Xingyu Zhu, Junkang Wu, Xiang Wang, Xiangnan He
The paper introduces Personalized Group Relative Policy Optimization (P‑GRPO), a new alignment framework for large language models that separates advantage estimation from immediate batch statistics. By normalizing advantages using preference‑group‑specific reward histories instead of the concurrent generation group, P‑GRPO maintains contrastive signals for distinct user preferences. Experiments across various tasks show that P‑GRPO converges faster and yields higher rewards than standard GRPO, improving alignment with heterogeneous human preferences while preserving general capabilities.
By Jialu Wang, Heinrich Peters, Asad A. Butt, Navid Hashemi, Alireza Hashemi, Pouya M. Ghari, Joseph Hoover, James Rae, Morteza Dehghani