AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

PPI Candidate Ranking: Large-Scale Evaluation of a Domain Knowledge–Guided Pipeline

Maria Emilia Russo, Federico Di Valerio, Alessia Borghini, Alessio Ragno, Roberto Capobianco

ICML 2026 · Large language models › Retrieval and knowledge · approach: Domain-specific QA datasets · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace raw prediction probabilities with interpretability-guided retrieval over active embedding regionsD-SCRIPT · STRING v11->v12Recall@10: 0.0124 -> 0.2641Improved
replace raw prediction probabilities with interpretability-guided retrieval over active embedding regionsTopsy-Turvy · STRING v11->v12Recall@10: 0.0117 -> 0.1106Improved
replace raw prediction probabilities with interpretability-guided retrievalD-SCRIPT · STRING v11->v12Average rank: 482.86 -> 239.77Improved
apply interpretability-guided retrieval instead of D-SCRIPT probability rankingD-SCRIPT · PiNUIRediscovery ratio: 0.0080 -> 0.3849Improved
apply interpretability-guided retrieval instead of D-SCRIPT probability rankingD-SCRIPT · PiNUISuccess@5 and Average rank: Success@5 0.010204 -> 0.000000; Avg rank 86.50 -> 924.78Hurt
re-rank top-10 candidates with fine-tuned PubMedBERT cross-encoder (ablation)Interpretability-guided retrieval (D-SCRIPT) · STRING v12maintain-or-improve rate vs cosine: 75.5%Improved
re-rank with D-SCRIPT interaction score (ablation)Interpretability-guided retrieval (D-SCRIPT) · STRING v12maintain-or-improve rate vs cosine: 63.0%Improved
re-rank with pDockQ structural score from SpeedPPI (ablation)Interpretability-guided retrieval (D-SCRIPT) · STRING v12maintain-or-improve rate vs cosine: 47.2%Hurt
re-rank with lightweight annotation overlaps (TF-IDF, token, location, key term) (ablation)Interpretability-guided retrieval (D-SCRIPT) · STRING v12maintain-or-improve rate vs cosine: up to ~70%Improved
re-rank with BioBERT or BioMedRoBERTa bi-encoders (ablation)Interpretability-guided retrieval (D-SCRIPT) · STRING v12maintain-or-improve rate vs cosine: 55.8 and 56.1Mixed
choose Topsy-Turvy over D-SCRIPT as backbone for retrieval (ablation)Interpretability-guided retrieval · STRING v11->v12early-rank precision, MAP, nDCG, success: Recall@10 0.1106 vs 0.2641Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy