This page is still being polished. If you have thoughts, please share them via the feedback form.
Data on this page is preliminary and may change. Please do not share or cite these figures publicly.
Non-Model mitigations not clearly fitting above categories.
Also in Non-Model
When the LLM possesses information pertaining to a specific prompt, enhancing the decoding strategy is a promising choice for mitigating hallucinations.
Typically, in contrast to conventional nucleus sampling (i.e., top-p sampling) used by the decoding procedure, factualnucleus sampling [113] gradually decays the value of p at each step of generation, as the generated content will become increasingly determined as the generation proceeds. Inspired by that the generation probability of a correct answer tends to incrementally rise from the lower layers to the higher layers, DoLa [282] computes the distribution of the next token based on the contrast between logits in a higher layer and that in a lower layer. After identifying a set of attention heads capable of eliciting the correct answer, ITI [283] intervenes with these selected attention heads. Motivated by that the contrasts between expert and amateur LMs can signal which generated text is better, Contrastive Decoding (CD) [284] is proposed to exploit such contrasts to guide the decoding process. In terms of the sycophancy issue, subtracting a sycophancy steering vector at the hidden layers can help reduce LLMs’ sycophantic tendency [285]. For the case that LLMs fail to exploit external knowledge introduced in the context, context-aware decoding (CAD) [180] is proposed to encourage LLMs to trust the input context if relevant input context is provided.
Reasoning
Runtime decoding strategies shape output generation to reduce hallucinations without modifying model weights.
Mitigation in Input Modules
Mitigating the threat posed by the input module presents a significant challenge for LLM developers due to the diversity of the harmful inputs and adversarial prompts [209], [210].
1.2.1 Guardrails & FilteringMitigation in Input Modules > Defensive Prompt Design
Directly modifying the input prompts is a viable approach to steer the behavior of the model and foster the generation of responsible outputs. This method integrates contextual information or constraints in the prompts to provide background knowledge and guidelines while generating the output [22].
1.2.1 Guardrails & FilteringMitigation in Input Modules > Malicious Prompt Detection
Different from the methods of designing defensive prompts to preprocess the input, the malicious prompt detection method aims to detect and filter out the harmful prompts through the input safeguard
1.2.1 Guardrails & FilteringMitigation in Language Models
This section delves into mitigating risks associated with models, encompassing privacy preservation, detoxification and debiasing, mitigation of hallucinations, and defenses against model attacks.
1.1 ModelMitigation in Language Models > Privacy Preserving
Privacy leakage is a crucial risk of LLMs, since the powerful memorization and association capabilities of LLMs raise the risk of revealing private information within the training data. Researchers are devoted to designing privacypreserving frameworks in LLMs [226], [227], aiming to safeguard sensitive PII from possible disclosure during humanmachine conservation
1.1 ModelMitigation in Language Models > Detoxifying and Debiasing
To reduce the toxicity and bias of LLMs, prior efforts mainly focus on enhancing the quality of training data and conducting safety training.
1.1 ModelRisk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems
Cui, Tianyu; Wang, Yanling; Fu, Chuanpu; Xiao, Yong; Li, Sijia; Deng, Xinhao; Liu, Yunpeng; Zhang, Qinglin; Qiu, Ziyi; Li, Peiyang; Tan, Zhixing; Xiong, Junwu; Kong, Xinyu; Wen, Zujie; Xu, Ke; Li, Qi (2024)
Despite their impressive capabilities, large lan- guage models (LLMs) have been observed to generate responses that include inaccurate or fabricated information, a phenomenon com- monly known as “hallucination”. In this work, we propose a simple Induce-then-Contrast De- coding (ICD) strategy to alleviate hallucina- tions. We first construct a factually weak LLM by inducing hallucinations from the original LLMs. Then, we penalize these induced hallu- cinations during decoding to enhance the fac- tuality of the generated content. Concretely, we determine the final next-token predictions by amplifying the predictions from the orig- inal model and downplaying the induced un- truthful predictions via contrastive decoding. Experimental results on both discrimination- based and generation-based hallucination eval- uation benchmarks, such as TruthfulQA and FACTSCORE, demonstrate that our proposed ICD methods can effectively enhance the factu- ality of LLMs across various model sizes and families. For example, when equipped with ICD, Llama2-7B-Chat and Mistral-7B-Instruct achieve performance comparable to ChatGPT and GPT4 on TruthfulQA, respectively.
Build and Use Model
Training, fine-tuning, and integrating the AI model
Developer
Entity that creates, trains, or modifies the AI system
Manage
Prioritising, responding to, and mitigating AI risks
