Large language models (LLMs) are increasingly used in information retrieval (IR) pipelines as relevance judges and re-rankers. Yet most analyses remain output-centric, evaluating generated labels o...
<p><strong>Tác giả:</strong> Pietro Bernardelle, Samaneh Mohtadi, Stefano Civelli, Joel Mackenzie, Gianluca Demartini</p>
<p>Large language models (LLMs) are increasingly used in information retrieval (IR) pipelines as relevance judges and re-rankers. Yet most analyses remain output-centric, evaluating generated labels or scores while offering limited insight into how relevance is represented inside the model. In this work, we study whether query-document (q-d) relevance is linearly decodable from residual-stream activations in instruction-tuned LLMs, how this signal compares with generated relevance judgments, and...</p>
<p>📄 DOI: <a href="https://doi.org/https://doi.org/10.1145/3799682.3841097" target="_blank" rel="noopener">https://doi.org/10.1145/3799682.3841097</a></p>
<p>📖 <a href="https://doi.org/https://doi.org/10.1145/3799682.3841097" target="_blank" rel="noopener">Xem bài báo đầy đủ</a></p>
Bài viết được biên soạn lại bởi LIFVietnam. Nguồn tham khảo: OpenAlex
