2000 character limit reached
Collaborative decoding of critical tokens for boosting factuality of large language models (2402.17982v1)
Published 28 Feb 2024 in cs.CL
Abstract: The most common training pipeline for LLMs includes pretraining, finetuning and aligning phases, with their respective resulting models, such as the pretrained model and the finetuned model. Finetuned and aligned models show improved abilities of instruction following and safe generation, however their abilities to stay factual about the world are impacted by the finetuning process. Furthermore, the common practice of using sampling during generation also increases chances of hallucination. In this work, we introduce a collaborative decoding framework to harness the high factuality within pretrained models through the concept of critical tokens. We first design a critical token classifier to decide which model to use for the next token, and subsequently generates the next token using different decoding strategies. Experiments with different models and datasets show that our decoding framework is able to reduce model hallucination significantly, showcasing the importance of the collaborative decoding framework.
- Training a helpful and harmless assistant with reinforcement learning from human feedback. ArXiv preprint, abs/2204.05862.
- Leo Breiman. 1996. Bagging predictors. Machine learning, 24:123–140.
- A survey on evaluation of large language models. ACM Transactions on Intelligent Systems and Technology.
- Dola: Decoding by contrasting layers improves factuality in large language models.
- Scaling instruction-finetuned language models. arXiv preprint arXiv:2210.11416.
- COLD: A benchmark for Chinese offensive language detection. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 11580–11599, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.
- Chain-of-verification reduces hallucination in large language models. arXiv preprint arXiv:2309.11495.
- Retrieval-augmented generation for large language models: A survey.
- A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. arXiv preprint arXiv:2311.05232.
- Few-shot learning with retrieval augmented language models. arXiv preprint arXiv:2208.03299.
- Mistral 7b. arXiv preprint arXiv:2310.06825.
- TriviaQA: A large scale distantly supervised challenge dataset for reading comprehension. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1601–1611. Association for Computational Linguistics.
- Natural questions: a benchmark for question answering research. Transactions of the Association of Computational Linguistics.
- Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing Systems, 33:9459–9474.
- The dawn after the dark: An empirical study on factuality hallucination in large language models.
- Halueval: A large-scale hallucination evaluation benchmark for large language models. In The 2023 Conference on Empirical Methods in Natural Language Processing.
- Inference-time intervention: Eliciting truthful answers from a language model. arXiv preprint arXiv:2306.03341.
- Contrastive decoding: Open-ended text generation as optimization. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).
- Truthfulqa: Measuring how models mimic human falsehoods. arXiv preprint arXiv:2109.07958.
- Mitigating the alignment tax of rlhf.
- Query rewriting for retrieval-augmented large language models. arXiv preprint arXiv:2305.14283.
- Factscore: Fine-grained atomic evaluation of factual precision in long form text generation. arXiv preprint arXiv:2305.14251.
- Sean O’Brien and Mike Lewis. 2023. Contrastive decoding improves reasoning in large language models. ArXiv preprint, abs/2309.09117.
- OpenAI. 2022. Introducing chatgpt. https://openai.com/blog/chatgpt.
- OpenAI. 2023. GPT-4 technical report. ArXiv preprint, abs/2303.08774.
- David Opitz and Richard Maclin. 1999. Popular ensemble methods: An empirical study. Journal of artificial intelligence research, 11:169–198.
- Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35:27730–27744.
- The troubling emergence of hallucination in large language models - an extensive definition, quantification, and prescriptive remediations. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 2541–2573, Singapore. Association for Computational Linguistics.
- A thorough examination of decoding methods in the era of llms. arXiv preprint arXiv:2402.06925.
- Retrieval augmentation reduces hallucination in conversation. arXiv preprint arXiv:2104.07567.
- Fine-tuning language models for factuality. arXiv preprint arXiv:2311.08401.
- Llama 2: Open foundation and fine-tuned chat models.
- Survey on factuality in large language models: Knowledge, retrieval and domain-specificity. arXiv preprint arXiv:2310.07521.
- Hallucination is inevitable: An innate limitation of large language models. arXiv preprint arXiv:2401.11817.
- Llm lies: Hallucinations are not bugs, but features as adversarial examples. arXiv preprint arXiv:2310.01469.
- Chain-of-note: Enhancing robustness in retrieval-augmented language models. arXiv preprint arXiv:2311.09210.
- SafeConv: Explaining and correcting conversational unsafe behavior. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 22–35, Toronto, Canada. Association for Computational Linguistics.
- Alleviating hallucinations of large language models through induced hallucinations. arXiv preprint arXiv:2312.15710.
- Verify-and-edit: A knowledge-enhanced chain-of-thought framework. arXiv preprint arXiv:2305.03268.
- Texygen: A benchmarking platform for text generation models. In The 41st international ACM SIGIR conference on research & development in information retrieval, pages 1097–1100.
Collections
Sign up for free to add this paper to one or more collections.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
