Papers with DPO training

8 papers
FocalPO: Enhancing Preference Optimizing by Focusing on Correct Preference Rankings (2025.acl-short)

Copied to clipboard

Challenge: Efficient preference optimization algorithms such as Direct Preference Optimization (DPO) have become a popular approach in aligning large language models with human preferences.
Approach: They propose a preference optimization variant that instead down-weighs misranked preference pairs and prioritizes enhancing the model’s understanding of pairs that it can already rank correctly.
Outcome: The proposed model outperforms DPO on benchmarks like Alpaca Eval 2.0 and Arena-Hard using mistral-base-7B and Llama-3-Instruct-8B with the introduced hyperparameter fixed.
Judge as A Judge: Improving the Evaluation of Retrieval-Augmented Generation through the Judge-Consistency of Large Language Models (2025.findings-acl)

Copied to clipboard

Challenge: Existing evaluation metrics cannot fairly evaluate the outputs of RAG models during training and evaluation.
Approach: They propose a method which prompts LLMs to generate different judgments based on various combinations of judgment dimensions and utilizes the judge-consistency to evaluate these judgments.
Outcome: The proposed method generates more accurate evaluations for RAG models across different RAG model and datasets.
Alignment with Fill-In-the-Middle for Enhancing Code Generation (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for generating test cases with limited training data are not reliable and may be counterproductive.
Approach: They propose a method that splits code snippets into smaller, granular blocks, creating more diverse DPO pairs from the same test cases.
Outcome: The proposed approach shows significant improvements in code generation tasks on benchmark datasets such as HumanEval (+), MBPP (+), and APPS.
Ambiguity Awareness Optimization: Towards Semantic Disambiguation for Direct Preference Optimization (2025.emnlp-main)

Copied to clipboard

Challenge: Direct Preference Optimization (DPO) is a widely used reinforcement learning from human feedback (RLHF) method across various domains.
Approach: They propose an approach that automatically re-weights ambiguous content to reduce ambiguities by calculating semantic similarity from preference pairs.
Outcome: The proposed approach outperforms state-of-the-art approaches in performance across multiple model scales and widely adopted benchmark datasets.
Towards Safety Reasoning in LLMs: AI-agentic Deliberation for Policy-embedded CoT Data Creation (2025.findings-acl)

Copied to clipboard

Challenge: Safety reasoning paradigms require high-quality policy-embedded chain-of-thought datasets . generating such data through human annotations is prohibitively expensive .
Approach: They propose AIDSAFE: Agentic Iterative Deliberation for Safety Reasoning . AIDS AFE leverages multi-agent deliberation to iteratively expand reasoning on safety policies .
Outcome: The proposed model improves policy adherence and reasoning quality while maintaining acceptable utility and over-refusal accuracy.
Topology-Enhanced Alignment for Large Language Models: Trajectory Topology Loss and Topological Preference Optimization (2026.findings-acl)

Copied to clipboard

Challenge: Existing algorithms for supervised fine-tuning and reinforcement learning from human feedback (RLHF) do not constrain how hidden states move from a user prompt to an answer.
Approach: They propose a topology-enhanced alignment framework that regularizes these trajectories using 0-dimensional persistent homology.
Outcome: The proposed framework regularizes semantic trajectory in hidden space using 0-dimensional persistent homology.
Filtered Direct Preference Optimization (2024.emnlp-main)

Copied to clipboard

Challenge: Existing studies on the impact of RLHF on text quality have focused on reward-model-free RL.
Approach: They propose an extension of direct preference optimization to improve model performance by analyzing the quality of the preference dataset.
Outcome: The proposed method improves the performance of models optimized with DPO over those optimized with reward-model-based RLHF.
Verified Critical Step Optimization for LLM Agents (2026.findings-acl)

Copied to clipboard

Challenge: Critical Step Optimization (CSO) focuses preference learning on verified critical steps where alternative actions demonstrably flip task outcomes from failure to success.
Approach: They propose a method which focuses preference learning on verified critical steps where alternative actions demonstrably flip task outcomes from failure to success.
Outcome: The proposed method outperforms the existing methods on GAIA-Text-103 and XBench-DeepSearch while requiring supervision at only 16% of trajectory steps.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations