AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Empty Shelves or Lost Keys? Recall Is the Bottleneck for Parametric Factuality

Nitay Calderon, Eyal Ben-David, Zorik Gekhman, Eran Ofek, Gal Yona

ICML 2026 · Large language models › Retrieval and knowledge · approach: Knowledge editing in LLMs · 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
enable thinkingGemini-3-Pro · WikiProfile (rare facts)recall gain (points): 20.1 vs 11.3 for popular factsImproved
enable thinkingGPT-5 · WikiProfile (reverse questions)recall gain (points): 19 vs 12 for direct questionsImproved
enable thinkingfrontier LLMs · WikiProfile (popularity gap)recall gap (points): 21.4 -> 12.5Improved
enable thinkingfrontier LLMs · WikiProfile (directionality gap)recall gap (points): 9 -> 2Improved
enable thinking on encoded vs non-encoded factsthinking-optimized LLMs · WikiProfilefacts recovered (%): 40-65 vs 5-15Improved
scale Gemma3 from 1B to 27BGemma3 · WikiProfileencoding failures (%): 85 -> 23Improved
scale Gemma3 model sizeGemma3 · WikiProfilerecall failure share (%): peaking at 33 (with thinking) and 40 (without)Hurt
switch from closed-book generation to multiple-choice verification for reverse questions13 LLMs · WikiProfilereverse vs direct recall: reverse easier for 9 models, similar for 4Improved
use reverse instead of direct questions (closed-book)GPT-5 · WikiProfiledirect recall (%): 82.9 -> 74Hurt
use rare instead of popular factsGemini-3-Pro · WikiProfiledirect recall among encoded (%): 84.6 -> 63.3Hurt
use rare instead of popular factsGemini-3-Pro · WikiProfileencoding rate (%): 99.5 -> 94.5Hurt
high-verbatim vs natural phrasing13 LLMs · WikiProfileperformance distribution: no significant effects in 104 testsNo clear effect

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