Reverse Preference Optimization for Complex Instruction Following (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for identifying and evaluating preference pairs with multiple constraints are noisy. |
| Approach: | They propose a method that dynamically reverses constraints to ensure the chosen response is perfect. |
| Outcome: | The proposed method reduces noise in preference pairs by reversing constraints to ensure the chosen response is perfect. |
Similar Papers
IOPO: Empowering LLMs with Complex Instruction Following via Input-Output Preference Optimization (2025.acl-long)
Copied to clipboard
| Challenge: | Existing algorithms to improve the ability of LLMs to follow complex instructions are lacking. |
| Approach: | They propose a benchmark to improve the ability to follow complex instructions by using a IOPO alignment method to take input and output preference into consideration. |
| Outcome: | The proposed algorithm shows 8.15%, 2.18% improvements on in-domain data and 5.91%, 2.83% on out-of-domain datasets compared to SFT and DPO respectively. |
MPPO: Multi Pair-wise Preference Optimization for LLMs with Arbitrary Negative Samples (2025.coling-main)
Copied to clipboard
Shuo Xie, Fangzhi Zhu, Jiahui Wang, Lulu Wen, Wei Dai, Xiaowei Chen, Junxiong Zhu, Kai Zhou, Bo Zheng
| Challenge: | Existing preference optimization methods such as DPO and KTO are inherently derived from PPO, requiring a reference model that adds GPU memory resources and relies heavily on abundant preference data. |
| Approach: | They propose an algorithm that leverages the average likelihood of model responses to fit the reward function and maximizes the utilization of preference data. |
| Outcome: | The proposed algorithm outperforms DPO, ORPO, and SimPO on MT-Bench and Arena-Hard. |
What Do LLMs Learn First? Asymmetric Learning Dynamics of Input Complexity and Output Ambiguity in Preference Alignment (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods treat all preference pairs uniformly during training. |
| Approach: | They propose a training framework that maintains separate, adaptive pacing schedules for each dimension. |
| Outcome: | The proposed training framework outperforms curriculum baselines by 2.1% and 0.21 points . it achieves 42.3% length-controlled win rate on AlpacaEval 2.0 and 7.66 on MT-Bench . |
sDPO: Don’t Use Your Data All at Once (2025.coling-industry)
Copied to clipboard
| Challenge: | Large language models (LLMs) are increasingly requiring precision and accuracy in alignment tuning. |
| Approach: | They propose a stepwise DPO technique that partitions available preference datasets incrementally rather than utilizing entire dataset simultaneously. |
| Outcome: | The proposed technique improves the accuracy of reference models and the overall performance of the final model. |
Enhancing Alignment using Curriculum Learning & Ranked Preferences (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Direct Preference Optimization (DPO) is an effective technique that leverages pairwise preference data to align LLMs to human preferences. |
| Approach: | They propose to use pairwise preference data to create multiple preference pairs for a given prompt. |
| Outcome: | The proposed method outperforms standard DPO on MTbench, Vicuna bench, and WizardLM with a score of 7.43 on the test sets. |
Reducing Hallucinations in LLMs via Factuality-Aware Preference Learning (2026.findings-acl)
Copied to clipboard
| Challenge: | Preference alignment methods can reinforce hallucinations when preference judgments reward fluency and confidence over factual correctness. |
| Approach: | They propose a method that corrects misordered preference pairs and adds a factuality-aware margin to emphasize pairs with clear correctness differences. |
| Outcome: | The proposed method improves factuality and reduces hallucination rates across seven open-weight LLMs. |
Weights-Rotated Preference Optimization for Large Language Models (2025.emnlp-main)
Copied to clipboard
Chenxu Yang, Ruipeng Jia, Mingyu Zheng, Naibin Gu, Zheng Lin, Siyuan Chen, Weichong Yin, Hua Wu, Weiping Wang
| Challenge: | Existing methods to align large language models with high reward hacking are limited by the complexity of the parameter space and the complexity. |
| Approach: | They propose a weights-rotated preference optimization algorithm that constrains the output layer logits with the KL divergence inherited from DPO and fine-tunes the intermediate hidden states. |
| Outcome: | The proposed algorithm achieves a 3.27-point improvement on AlpacaEval 2 and surpasses the best baseline by 6.2 to 7.5 points on MT-Bench with merely 0.015% of the trainable parameters. |
Insights into Alignment: Evaluating DPO and its Variants Across Multiple Tasks (2025.acl-srw)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) excel in math reasoning problemsolving, text generation, summarization, creative writing, among other tasks. |
| Approach: | They evaluate Direct Preference Optimization and its variants for aligning Large Language Models with human preferences. |
| Outcome: | The proposed alignment methods achieve near-optimal performance even with smaller subsets of training data. |
Eliminating Biased Length Reliance of Direct Preference Optimization via Down-Sampled KL Divergence (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies attributed verbosity to biased labels, but new research shows that DPO can be effective in mitigating verboses. |
| Approach: | They propose to use a method to reduce the amount of verbosity in LLMs by using a downsampling approach. |
| Outcome: | The proposed approach overcomes the problem of verbosity by reducing the length reliance of the proposed algorithm. |
Comparing Bad Apples to Good Oranges Aligning Large Language Models via Joint Preference Optimization (2025.findings-acl)
Copied to clipboard
| Challenge: | Recent studies have shown that acquiring human preferences by comparing generations is not effective for large language models. |
| Approach: | They propose a preference optimization objective that elicits preferences jointly over the instruction-response pairs. |
| Outcome: | The proposed approach outperforms prior preference optimizations by 5.2% and 3.3% in summarization and open-ended dialogue datasets. |