CONGRAD: Conflicting Gradient Filtering for Multilingual Preference Alignment (2026.eacl-long)
Copied to clipboard
Jiangnan Li, Thuy-Trang Vu, Christian Herold, Amirhossein Tebbifakhr, Shahram Khadivi, Gholamreza Haffari
| Challenge: | Naive joint training of large language models can suffer from negative interference. |
| Approach: | They propose a filtering method that aggregates cross-lingually beneficial gradients and filters for those with high cross-linguistic affinity. |
| Outcome: | The proposed method outperforms baselines in both seen and unseen languages with minimal alignment tax. |
Similar Papers
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. |
Understanding Cross-Lingual Alignment—A Survey (2024.findings-acl)
Copied to clipboard
| Challenge: | Cross-lingual alignment is the meaningful similarity of representations across languages in multilingual language models. |
| Approach: | They propose a taxonomy of methods to improve cross-lingual alignment . they argue that an effective trade-off between language-neutral and language-specific information is key . |
| Outcome: | The proposed methods can be applied to encoder models and encoder-decoder-only models . they show that language-neutral and language-specific information is key . |
Implicit Cross-Lingual Rewarding for Efficient Multilingual Preference Alignment (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to align English LLMs with human preferences rely on expensive human annotations or advanced multilingual preference alignment models. |
| Approach: | They propose a method that captures learned preferences from English models by implicit rewards . they annotate preference relations in cross-lingual instruction-following pairs using English . |
| Outcome: | The proposed approach captures learned preferences from well-aligned English models by implicit rewards and transfers them to other languages through iterative training. |
CM-Align: Consistency-based Multilingual Alignment for Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Current large language models (LLMs) show a significant performance gap in alignment between English and other languages. |
| Approach: | They propose a consistency-based method to construct high-quality multilingual preference data for improving multilingual alignment. |
| Outcome: | The proposed method is based on three LLMs and three common tasks and shows that it performs better than current methods. |
MPO: Multilingual Safety Alignment via Reward Gap Optimization (2025.acl-long)
Copied to clipboard
Weixiang Zhao, Yulin Hu, Yang Deng, Tongtong Wu, Wenxuan Zhang, Jiahe Guo, An Zhang, Yanyan Zhao, Bing Qin, Tat-Seng Chua, Ting Liu
| Challenge: | Existing preference learning methods for safety alignment are monolingual and struggle with noisy multilingual data. |
| Approach: | They propose a multilingual reward gaP optimization approach that leverages the well-aligned safety capabilities of the dominant language to improve safety alignment across multiple languages. |
| Outcome: | Extensive experiments on three LLMs, LLaMA-3.1, Gemma-2 and Qwen2.5, validate MPO’s efficacy in multilingual safety alignment without degrading general multilingual utility. |
Aligning LLMs with Individual Preferences via Interaction (2025.coling-main)
Copied to clipboard
| Challenge: | Existing studies on LLMs alignment focus on generalizing their behavior to generalized values such as helpfulness, harmlessness, and honesty. |
| Approach: | They train large language models to "interact to align" to implicitly infer user preferences . they use a multi-turn preference dataset to generate a personalized alignment . |
| Outcome: | The proposed method enables dynamic, personalized alignment via interaction with a multi-turn preference dataset. |
LaCoMSA: Language-Consistency Multilingual Self-Alignment with Latent Representation Rewarding (2026.eacl-long)
Copied to clipboard
| Challenge: | Existing multilingual alignment methods mitigate these issues but rely on external supervision, such as translation systems or English-biased signal. |
| Approach: | They propose a preference optimization framework that leverages an LLM’s own latent representations as intrinsic supervision signals and rewards lower-resource language outputs based on their alignment with high-resourced (English) counterparts in the "semantic hub". |
| Outcome: | The proposed framework improves a Llama 3 8B model multilingual win rates by up to 6.8% absolute (55.0% relative) on X-AlpacaEval and achieves consistent gains across benchmarks and models. |
Gradient-Adaptive Policy Optimization: Towards Multi-Objective Alignment of Large Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Reinforcement learning from human feedback (RLHF) has emerged as a powerful technique for aligning large language models (LLMs) with human preferences. |
| Approach: | They propose a novel algorithm that uses multiple-gradient descent to optimize LLMs with diverse preferences to maximize trade-offs between objectives. |
| Outcome: | The proposed approach incorporates user preferences across different objectives and achieves Pareto solutions that better align with the user’s specific needs. |
From Unaligned to Aligned: Scaling Multilingual LLMs with Multi-Way Parallel Corpora (2025.emnlp-main)
Copied to clipboard
| Challenge: | Experiments show that models trained on multi-way parallel data outperform those trained on unaligned data. |
| Approach: | They propose a large-scale, high-quality multi-way parallel corpus based on TED Talks that spans 113 languages with up to 50 languages aligned in parallel. |
| Outcome: | The proposed model outperforms models trained on unaligned multilingual data on six multilingual benchmarks. |
Gradient-based Gradual Pruning for Language-Specific Multilingual Neural Machine Translation (2023.emnlp-main)
Copied to clipboard
| Challenge: | Multilingual neural machine translation suffers from performance degradation in high-resource languages compared to bilingual counterparts. |
| Approach: | They propose a gradient-based gradual pruning technique for multilingual neural machine translation that allows for partial parameter sharing across language pairs to alleviate interference. |
| Outcome: | The proposed approach yields a notable performance gain on IWSLT and WMT datasets. |