Papers by Ravinarayana Adkathimar
SelfElicit: Your Language Model Secretly Knows Where is the Relevant Evidence (2025.acl-long)
Copied to clipboard
| Challenge: | Recent studies have found that Language Models struggle to fully comprehend and utilize key evidence from the context. |
| Approach: | They propose an inference-time approach that helps LMs focus on key contextual evidence through self-guided explicit highlighting. |
| Outcome: | The proposed method improves on multiple evidence-based QA tasks while maintaining computational efficiency. |