Papers with sampling-based methods
Enhancing Hallucination Detection via Future Context (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are widely used to generate plausible text on online platforms, without revealing the generation process. |
| Approach: | They propose a framework for detection of hallucinations in black-box generators by analyzing future contexts. |
| Outcome: | The proposed framework improves on existing methods and demonstrates that it is feasible to integrate it with other models. |
Addressing Both Statistical and Causal Gender Fairness in NLP Models (2024.findings-naacl)
Copied to clipboard
| Challenge: | Statistical fairness stipulates equivalent outcomes for all protected groups, whereas causal fairness prescribes that a model makes the same prediction for an individual regardless of their protected characteristics. |
| Approach: | They propose to use statistical and causal debiasing methods to reduce gender bias in NLP models. |
| Outcome: | The proposed methods reduce gender bias measured by the targeted metric, but not on other bias metrics. |
Generating EDU Extracts for Plan-Guided Summary Re-Ranking (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods to generate summary candidates for re-ranking produce redundant, and often low quality, content. |
| Approach: | They propose a method to generate candidates for re-ranking that addresses these issues by grounding each abstract on its own unique content plan and creating distinct plan-guided abstracts using a model's top beam. |
| Outcome: | The proposed method outperforms baseline decoding methods on CNN, NYT, and Xsum and shows that prompting GPT-3 to follow EDU plans outperformed sampling-based methods by 1.05 points. |
FactSelfCheck: Fact-Level Black-Box Hallucination Detection for LLMs (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing methods to detect hallucinated content are limited by their tendency to generate factual errors. |
| Approach: | They propose a black-box sampling-based method that enables fine-grained fact-level detection by representing text as interpretable knowledge graphs consisting of facts in the form of triples. |
| Outcome: | The proposed method improves hallucination correction by 35.5% compared to baseline methods while sentence-level SelfCheckGPT yields only 10.6% improvement. |
Rethinking Model Selection and Decoding for Keyphrase Generation with Pre-trained Sequence-to-Sequence Models (2023.emnlp-main)
Copied to clipboard
| Challenge: | Keyphrase generation is a longstanding task in NLP with widespread applications. |
| Approach: | They propose a likelihood-based decode-select algorithm for seq2seq PLMs that improves greedy search by an average of 4.7% semantic F1 across five datasets. |
| Outcome: | The proposed algorithm improves greedy search by an average of 4.7% semantic F1 across five datasets. |