# life: detecting hallucinations (in large language models) using semantic entropy.

<< Large language model (LLM) systems, such as ChatGPT or Gemini, can show impressive reasoning and question-answering capabilities but often ‘hallucinate’ false outputs and unsubstantiated answers. >>

<< Here (AA) develop new methods grounded in statistics, proposing entropy-based uncertainty estimators for LLMs to detect a subset of hallucinations— confabulations— which are arbitrary and incorrect generations. (Their) method addresses the fact that one idea can be expressed in many ways by computing uncertainty at the level of meaning rather than specific sequences of words. >>

Their method << works across datasets and tasks without a priori knowledge of the task, requires no task-specific data and robustly generalizes to new tasks not seen before. By detecting when a prompt is likely to produce a confabulation, helps users understand when they must take extra care with LLMs and opens up new possibilities for using LLMs that are otherwise prevented by their unreliability. >>️️

Sebastian Farquhar, Jannik Kossen, et al. Detecting hallucinations in large language models using semantic entropy. Nature 630, 625–630. Jun 19, 2024.

https://www.nature.com/articles/s41586-024-07421-0

Also: ai (artificial intell) (bot), entropy, in https://www.inkgmr.net/kwrds.html 

Keywords: life, artificial intelligence,  LLMs, confabulations, uncertainty, hallucinations, entropy, semantic entropy

# gst: trade-off between coherence and dissipation for excitable phase oscillators.

<< Thermodynamic uncertainty relation (TUR) bounds coherence in stochastic oscillatory systems. In this paper, (AA) show that both dynamical and thermodynamic bounds play important roles for the excitable oscillators, e.g. neurons. >>

<< Excitable systems such as neurons have distinctive coherence features compared with other oscillators having no excitability. >>️

AA << combined the well-established results, i.e. the fluctuation of the ISI (inter-spike-interval) limited by 1/3 and the coherence resonance phenomenon, together with the TUR developed in recent years to investigate the coherence in the excitable phase oscillators. (AA) find quite different trade-off relation in the subthreshold (excitable) region and superthreshold (oscillatory) region, separated by the SNIC (saddle-node on an invariant circle) bound but meanwhile lower bounded by the TUR. Furthermore, (They) found that there is an optimal entropy production corresponding to the maximum coherence, which could serve as an alternative interpretation of the coherence resonance. It implies that more entropy production does not necessarily result in higher accuracy of currents. >>️

Chunming Zheng. Trade-off between coherence and dissipation for excitable phase oscillators. arXiv: 2412.16603v1 [cond-mat.stat-mech]. Dec 21, 2024.

https://arxiv.org/abs/2412.16603

Also: brain, entropy, dissipation, uncertainty, in https://www.inkgmr.net/kwrds.html 

Keywords: gst, brain, neurons, entropy, oscillators, excitable phase oscillators, coherence, dissipation, uncertainty