Papers with SimPO
Harnessing Negative Signals: Reinforcement Distillation from Teacher Data for LLM Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Recent advances in model distillation show that data from advanced reasoning models can effectively train smaller student models. |
| Approach: | They propose a method to use both positive and negative distilled reasoning traces to maximize LLM reasoning performance in offline settings. |
| Outcome: | The proposed model outperforms existing methods in the distillation context. |
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. |
SPO: Self Preference Optimization with Self Regularization (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing reference-free preference optimization methods exhibit higher training efficiency but are prone to overoptimization, leading to performance degradation. |
| Approach: | They propose a reference-free preference optimization method that replaces the logsigmoid loss function with a SiLU function to improve the model's performance. |
| Outcome: | The proposed method achieves 7% improvement over SimPO on AlpacaEval 2 and MT-Bench. |
ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning (2025.emnlp-main)
Copied to clipboard
Ziqing Qiao, Yongheng Deng, Jiali Zeng, Dong Wang, Lai Wei, Guanbo Wang, Fandong Meng, Jie Zhou, Ju Ren, Yaoxue Zhang
| Challenge: | Existing methods for fine-tuning-based compression suffer from verbose outputs, increasing computational overhead. |
| Approach: | They propose a framework to generate concise reasoning chains using Confidence Injection and Early Stopping. |
| Outcome: | The proposed framework reduces the length of the model by up to 50% while maintaining high task accuracy. |
A Practical Analysis of Human Alignment with *PO (2025.findings-naacl)
Copied to clipboard
| Challenge: | Prior research focused on identifying the best-performing method to varying hyperparameters . prior research focused primarily on a grid search, which can be impractical for general practitioners . |
| Approach: | They propose a preference optimization method that is more stable across hyperparameters and reduces the average response length. |
| Outcome: | The proposed method increases likelihood of achieving better results through various metrics, such as KL divergence and response length. |
Towards Bridging the Reward-Generation Gap in Direct Alignment Algorithms (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks, including instruction following, mathematical problem solving, and coding generation. |
| Approach: | They propose a method that truncates both preferred and dispreferred responses to match the shorter one’s length. |
| Outcome: | The proposed approach improves over standard implementations and achieves 11.8 points in AlpacaEval 2 and overall improvements across downstream tasks. |
Teaching an Old LLM Secure Coding: Localized Preference Optimization on Distilled Preferences (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods to improve security of LLM generated code are ineffective and lack localized regions of code. |
| Approach: | They propose a method for distilling a preference dataset of insecure and secure code pairs from frontier LLMs and a security reasoning that explains the issues and the fix. |
| Outcome: | The proposed method reduces code insecurity while improving overall code quality. |
Beyond the Leaderboard: Understanding Performance Disparities in Large Language Models via Model Diffing (2025.emnlp-main)
Copied to clipboard
| Challenge: | a recent study shows that benchmarking fails to explain why models outperform others . open-weight large language models have transformed the AI landscape . |
| Approach: | They use model diffing to analyze capability differences between Gemma-2-9b-it and SimPO-enhanced variants. |
| Outcome: | The proposed model diffing approach can provide fine-grained insights beyond leaderboard metrics . it can also help to identify model performance gaps, the authors say . |