From Evidence to Belief: A Bayesian Epistemology Approach to Language Models (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing studies on language models have focused on factual correctness and justification, but prior research has focused on the factual truth condition and justifier. |
| Approach: | They analyze language models’ responses and confidence using verbalized confidence, token probability, and sampling to examine their knowledge of Bayesian epistemology. |
| Outcome: | The language models that follow the Bayesian confirmation assumption with true evidence show varying performance depending on the degree of irrelevance, indicating they deviate from Bayes' assumptions. |
Similar Papers
Confidence Under the Hood: An Investigation into the Confidence-Probability Alignment in Large Language Models (2024.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are increasingly used in high-stakes areas such as healthcare, law, and education. |
| Approach: | They propose a concept of Confidence-Probability Alignment that connects an LLM’s internal confidence to the confidence conveyed in the model’s response when explicitly asked about its certainty. |
| Outcome: | The proposed model shows the strongest confidence-probability alignment across a wide range of tasks. |
How Can We Know When Language Models Know? On the Calibration of Language Models for Question Answering (2021.tacl-1)
Copied to clipboard
| Challenge: | Recent studies have shown that language models capture different types of knowledge regarding facts or commonsense knowledge. |
| Approach: | They examine how language models can be calibrated to make their confidence scores correlate better with the likelihood of correctness. |
| Outcome: | The proposed calibration methods improve confidence scores on QA tasks and improve accuracy. |
What Evidence Do Language Models Find Convincing? (2024.acl-long)
Copied to clipboard
| Challenge: | Current retrieval-augmented language models are tasked with subjective, contentious, and conflicting queries. |
| Approach: | They construct a dataset that pairs controversial queries with real-world evidence documents . they find current models rely heavily on relevance of a website to the query . |
| Outcome: | The proposed dataset pairs controversial queries with real-world evidence documents that contain different facts, arguments, and answers. |
Perceptions of Linguistic Uncertainty by Language Models and Humans (2024.emnlp-main)
Copied to clipboard
| Challenge: | Prior work has shown that humans are well-attuned to the use of uncertainty expressions, exhibiting population-level agreement in mapping these expressions to numerical responses. |
| Approach: | They propose to map linguistic expressions of uncertainty to numerical responses by using a theory of mind approach to understand the uncertainty of another agent. |
| Outcome: | The proposed model can map expressions to probabilistic responses in a human-like manner, but different behavior depending on whether a statement is actually true or false. |
A Survey of Confidence Estimation and Calibration in Large Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) have demonstrated impressive capabilities across a wide range of tasks in various domains, but they can be unreliable due to factual errors in their generations. |
| Approach: | They summarize recent advances in LLM confidence estimation and calibration and outline their main lessons learned. |
| Outcome: | The proposed methods can be used to assess the reliability of models and to calibrate them across tasks. |
Credible without Credit: Domain Experts Assess Generative Language Models (2023.acl-short)
Copied to clipboard
| Challenge: | ChatGPT has been criticized for its lack of accuracy and coherence . authors argue that language models could replace search engines and make college essays obsolete . |
| Approach: | a team of 10 domain experts conducts an initial assessment of language models using 100 expert-written questions. |
| Outcome: | The results show that language models are mixed in their accuracy. |
Deep Bayesian Learning and Understanding (C18-3)
Copied to clipboard
| Challenge: | COLING 2018 is a conference for researchers and practitioners working on machine learning and deep learning. |
| Approach: | a tutorial on machine learning and deep learning will be presented at COLING 2018 . the tutorial will focus on statistical models, deep neural networks, sequential learning and natural language understanding . |
| Outcome: | This tutorial will present the latest advances in deep Bayesian and sequential learning at COLING 2018 . |
Can Large Language Models Discern Evidence for Scientific Hypotheses? Case Studies in the Social Sciences (2024.lrec-main)
Copied to clipboard
| Challenge: | scholarly databases fail to aggregate, compare, contrast, and contextualize existing studies in service to a targeted research question. |
| Approach: | They propose to use large language models to discern evidence in support or refute of specific hypotheses based on abstracts. |
| Outcome: | The proposed method outperforms state-of-the-art methods and highlights opportunities for future research. |
Epistemology of Language Models: Do Language Models Have Holistic Knowledge? (2024.findings-acl)
Copied to clipboard
| Challenge: | et al., 2021) explores whether language models exhibit characteristics consistent with epistemological holism . authors examined the epistle of language models from the perspective of abduction, revision, and argument generation tasks. |
| Approach: | They examine whether language models exhibit characteristics consistent with epistemological holism . they created a scientific reasoning dataset and examined the epistology of language models . |
| Outcome: | The language models showed that they did not distinguish between core and peripheral knowledge, compared with other tasks. |
Factual Confidence of LLMs: on Reliability and Robustness of Current Estimators (2024.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) tend to be unreliable on fact-based answers. |
| Approach: | They propose a framework for comparing LLMs' confidence over fact-based answers with hidden-state probes that are more reliable than hidden-status probes. |
| Outcome: | The proposed methods show that hidden-state probes provide the most reliable confidence estimates despite requiring access to weights and supervision data. |