Papers2

#hit rate

Conv-FinRe: A Conversational and Longitudinal Benchmark for Utility-Grounded Financial Recommendation

This paper builds Conv-FinRe, a new test that checks if AI financial advisors give advice that fits a person’s true goals, not just what they clicked before.

#financial recommendation#utility-based evaluation#conversational benchmark

Not triaged yet

Reveal Hidden Pitfalls and Navigate Next Generation of Vector Similarity Search from Task-Centric Views

Beginner

Tingyang Chen, Cong Fu et al.Dec 15arXiv

The paper shows that judging vector search only by distance-based recall and speed can be very misleading for real tasks.

Not triaged yet