paper-with-me

홈 › Papers

Self-Judge: Selective Instruction Following with Alignment Self-Evaluation

2024-09-02 · Hai Ye, Hwee Tou Ng

Pre-trained large language models (LLMs) can be tailored to adhere to human instructions through instruction tuning. However, due to shifts in the distribution of test-time data, they may not always execute instructions accurately, potentially generating factual errors or misaligned content when acting as chat assistants. To enhance the reliability of LLMs in following instructions, we propose the study of selective instruction following, whereby the system declines to execute instructions if the anticipated response quality is low. We train judge models that can predict numerical quality scores for model responses. To address data scarcity, we introduce Self-J, a novel self-training framework for developing judge models without needing human-annotated quality scores. Our method leverages the model's inherent self-evaluation capability to extract information about response quality from labeled instruction-tuning data. It incorporates a gold reference answer to facilitate self-evaluation and recalibrates by assessing the semantic similarity between the response sample and the gold reference. During the training phase, we implement self-distillation as a regularization technique to enhance the capability of reference-free estimation. To validate alignment evaluation on general instruction-following tasks, we collect large-scale high-quality instructions from Hugging Face for model training and evaluation. Extensive experiments on five open-source models show that our method correlates much more with GPT-4 than strong baselines, e.g., supervised models distilled from GPT-4 and GPT-3.5-turbo. Our analysis shows our model's strong generalization across domains. Additionally, our judge models serve as good reward models, e.g., boosting WizardLM-13B-V1.2 from 89.17 to 92.48 and from 12.03 to 15.90 in version v1 and v2 of AlpacaEval respectively using best-of-32 sampling with our judge models.

📄 PDF Abstract BibTeX arXiv:2409.00935

Code (1)

nusnlp/Self-J 공식 구현 pytorch

Tasks

Instruction FollowingSemantic SimilaritySemantic Textual Similarity

Methods 이 논문이 사용한 방법론

Refunds@Expedia|||How do I get a full refund from Expedia? “How do I get a full refund from Expedia? How do I get a full refund from Expedia? – Call ☎️ +1-(888) 829 (0881) or +1-805-330-4056 or +1-805-330-4056 for Quick Help &…
15 Ways to Contact How can i speak to someone at Delta Airlines 설명 없음
Attention 설명 없음
Cosine Annealing Cosine Annealing is a type of learning rate schedule that has the effect of starting with a large learning rate that is relatively rapidly decreased to a minimum value before…
Absolute Position Encodings Absolute Position Encodings are a type of position embeddings for [Transformer-based models] where positional encodings are…
Label Smoothing Label Smoothing is a regularization technique that introduces noise for the labels. This accounts for the fact that datasets may have mistakes in them, so maximizing the…
Position-Wise Feed-Forward Layer 설명 없음
Residual Connection 설명 없음

Similar Papers 제목 키워드 기반

IF-RewardBench: Benchmarking Judge Models for Instruction-Following Evaluation

2026-03-05 · Bosi Wen, Yilin Niu, Cunxiang Wang, Xiaoying Ling 외 arxiv

Instruction-following is a foundational capability of large language models (LLMs), with its improvement hinging on scalable and accurate feedback from judge models. However, the reliability of current judge models in in…

Becoming Experienced Judges: Selective Test-Time Learning for Evaluators

2025-12-07 · Seungyeon Jwa, Daechul Ahn, Reokyoung Kim, Dongyeop Kang 외 arxiv

Automatic evaluation with large language models, commonly known as LLM-as-a-judge, is now standard across reasoning and alignment tasks. Despite evaluating many samples in deployment, these evaluators typically (i) treat…

Aligning Large Language Models by On-Policy Self-Judgment

2024-02-17 · Sangkyu Lee, Sungdong Kim, Ashkan Yousefpour, Minjoon Seo 외

Existing approaches for aligning large language models with human preferences face a trade-off that requires a separate reward model (RM) for on-policy learning. In this paper, we present a novel alignment framework, SEL…

Instruction Following

LLM-as-a-Tutor: Policy-Aware Prompt Adaptation for Non-Verifiable RL

2026-07-05 · Yujin Kim, Namgyu Ho, Sangmin Hwang, Joonkee Kim 외 hf

Reinforcement learning (RL) for non-verifiable instruction following increasingly relies on LLM judges with prompt-specific rubrics as reward signals. While recent methods adapt these rubrics to the evolving policy durin…

Reinforcement LearningInstruction Following

Better Instruction-Following Through Minimum Bayes Risk

2024-10-03 · Ian Wu, Patrick Fernandes, Amanda Bertsch, Seungone Kim 외

General-purpose LLM judges capable of human-level evaluation provide not only a scalable and accurate way of evaluating instruction-following LLMs but also new avenues for supervising and improving their performance. One…

Instruction Following