MIT AI Risk Repository · Risk Sub-Category · 02.09.03
Defective Decoding Process
Category: Hallucinations
Description
In general, LLMs employ the Transformer architecture [32] and generate content in an autoregressive manner, where the prediction of the next token is conditioned on the previously generated token sequence. Such a scheme could accumulate errors [105]. Besides, during the decoding process, top-p sampling [28] and top-k sampling [27] are widely adopted to enhance the diversity of the generated content. Nevertheless, these sampling strategies can introduce “randomness” [113], [136], thereby increasing the potential of hallucinations"
From Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024), as extracted by the MIT AI Risk Repository (CC BY 4.0).
Classification
- Domain
- 3. Misinformation
- Subdomain
- 3.1 False or misleading information
- Causal entity
- AI
- Intent
- Unintentional
- Timing
- Pre-deployment
Subdomain definition: AI systems that inadvertently generate or spread incorrect or deceptive information, which can lead to inaccurate beliefs in users and undermine their autonomy. Humans that make decisions based on false beliefs can experience physical, emotional or material harms
Real-world incidents in this subdomain
- Nonfiction Book 'The Future of Truth' Reportedly Included AI-Generated and Misattributed Quotations
- Claude Console Reportedly Generated Phantom Legal Quotations in Trump Layoffs Court Filing
- Purportedly AI-Enhanced Images of Iranian Women Protesters Were Reportedly Spread With Unverified Execution Claims
- South Africa Draft National AI Policy Reportedly Included Fictitious References Believed to Be AI Hallucinations
- Purportedly AI-Generated Image Reportedly Misled Daejeon Authorities Searching for Escaped Wolf Neukgu
- Gemini and Grok Reportedly Misidentified Authentic Minab School-Strike Graveyard Photo as Unrelated Disaster Imagery