Reducing LLM Hallucinations using Epistemic Neural Networks (2312.15576v1)
Abstract: Reducing and detecting hallucinations in LLMs is an open research problem. In this project, we attempt to leverage recent advances in the field of uncertainty estimation to reduce hallucinations in frozen LLMs. Epistemic neural networks have recently been proposed to improve output joint distributions for large pre-trained models. ENNs are small networks attached to large, frozen models to improve the model's joint distributions and uncertainty estimates. In this work, we train an epistemic neural network on top of the Llama-2 7B model combined with a contrastive decoding feature enhancement technique. We are the first to train an ENN for the next token prediction task and explore the efficacy of this method in reducing hallucinations on the TruthfulQA dataset. In essence, we provide a method that leverages a pre-trained model's latent embeddings to reduce hallucinations.
- Edm3: Event detection as multi-task text generation. ArXiv, abs/2305.16357, 2023. URL https://api.semanticscholar.org/CorpusID:258947208.
- Chuang et al. Dola: Decoding by contrasting layers improves factuality in large language models. In Submitted to The Twelfth International Conference on Learning Representations, 2023. URL https://openreview.net/forum?id=Th6NyL07na. under review.
- Context-ner : Contextual phrase generation at scale. 2021. URL https://api.semanticscholar.org/CorpusID:265039305.
- Instruction tuned models are quick learners. ArXiv, abs/2306.05539, 2023a. URL https://api.semanticscholar.org/CorpusID:259129868.
- Targen: Targeted data generation with large language models. ArXiv, abs/2310.17876, 2023b. URL https://api.semanticscholar.org/CorpusID:264555527.
- Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing Systems, 33:9459–9474, 2020.
- Contrastive decoding: Open-ended text generation as optimization, 2023.
- Visual instruction tuning, 2023.
- OpenAI. Gpt-4 technical report, 2023.
- Epistemic neural networks. arXiv preprint arXiv:2107.08924, 2021.
- Fine-tuning language models via epistemic neural networks. arXiv preprint arXiv:2211.01568, 2022a.
- The neural testbed: Evaluating joint predictions. Advances in Neural Information Processing Systems, 35:12554–12565, 2022b.
- Exploring the limits of transfer learning with a unified text-to-text transformer, 2023.
- Instructabsa: Instruction learning for aspect based sentiment analysis. ArXiv, abs/2302.08624, 2023. URL https://api.semanticscholar.org/CorpusID:257020097.
- BERT rediscovers the classical NLP pipeline. In Anna Korhonen, David Traum, and Lluís Màrquez (eds.), Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 4593–4601, Florence, Italy, July 2019. Association for Computational Linguistics. doi: 10.18653/v1/P19-1452. URL https://aclanthology.org/P19-1452.
- Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023.
- Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171, 2022.
- Chain-of-thought prompting elicits reasoning in large language models. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh (eds.), Advances in Neural Information Processing Systems, volume 35, pp. 24824–24837. Curran Associates, Inc., 2022. URL https://proceedings.neurips.cc/paper_files/paper/2022/file/9d5609613524ecf4f15af0f7b31abca4-Paper-Conference.pdf.
- Can llms express their uncertainty? an empirical evaluation of confidence elicitation in llms. arXiv preprint arXiv:2306.13063, 2023.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.