HN Thread Discusses Self-Referentiality in LLMs

Related stories

• Linguistic Illegibility Raises LLM Security Concerns

  A Hacker News discussion examines how linguistic illegibility—ambiguity and opacity in language—poses security risks for large language models. Participants debate the implications for AI safety, suggesting that unclear or deceptive language could be exploited to bypass model safeguards. The conversation highlights the need for robust defenses against such linguistic vulnerabilities.

  1 reports · 1 sources78 points · 29 commentslast month

• Paper Proposes Direct Semantic Communication Between LLMs via Cache-to-Cache

  A 2025 paper, discussed on Hacker News, proposes a method for direct semantic communication between large language models by transferring cache states, bypassing traditional text-based exchanges. The approach aims to enable more efficient and nuanced information sharing between models.

  1 reports · 1 sources107 points · 20 commentslast month

• Infinite-Parameter LLMs Generate and Adapt Weights from Live Data

  A Hacker News discussion highlights a research concept proposing infinite-parameter large language models that generate and adapt their weights from live data. The approach suggests a shift from static, pre-trained models to dynamic systems that continuously update based on incoming information.

  1 reports · 1 sources157 points · 43 commentslast month

• Hacker News Discussion: LLM Classification Viewed as Feature Engineering

  A Hacker News discussion argues that LLM classification tasks are fundamentally about feature engineering, offering a technically substantive perspective on model behavior. The discussion highlights how the approach to classification with LLMs can be understood through the lens of feature engineering, potentially influencing how researchers and engineers design and evaluate such systems.

  1 reports · 1 sources114 points · 24 commentslast month