Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval (2020 · 688 citations)
Dense retrieval (DR) encodes queries and documents as vectors and matches them by similarity. In practice it often lost to sparse BM25, especially on documents. The authors trace this weakness to how negatives are chosen during training. Using a variance-reduction view of SGD, they argue that random or in-batch…
EvolvR: Self-Evolving Pairwise Reasoning for Story Evaluation to Enhance Generation (2026 · 1 citation)
LLM judges remain weak at assessing stories. Prompting closed models is brittle, and fine-tuned open evaluators usually output scores without explicit reasoning. The authors propose EvolvR, a framework that teaches an open-source evaluator to reason over pairs of stories. The model works aspect by aspect, for example…
Triviality Corrected Endogenous Reward (2026)
Reinforcement fine-tuning for open-ended writing lacks verifiable rewards. It usually depends on costly proprietary LLM judges or on judge models trained with annotated preference data. The authors ask whether confidence-based endogenous rewards, which work for unsupervised RL in math, can serve writing instead. They…

Co-authors here

10 co-authors with a page, the most papers together first.

Venues

Where their works here were published, the most works first, with the editions of those that have a page.

Papers of our library cite their works 485 times (of the 149,439 papers whose bibliographies we read); 71 citing papers with a page here →.

Updated 1 November 2026