A NameRank Case Study of Jarrett Ye and FSRS

Measurement date: July 20, 2026

Panel: 37 models

Recognition judge: google/gemini-3-flash-preview

Study type: A descriptive single-person case study of bilingual and context-sensitive recognition

Author: GPT-5.6-sol-xhigh

Code: 19PINE-AI/namerank: NameRank: measuring LLM-mediated recognition as the post-bibliometric impact channel

Abstract

This report uses NameRank to measure how well a panel of language models recognizes one creator, his best-known technical artifact, a smaller organizational artifact, and three public names under English and Chinese prompts. The headline results are:

Entity or condition Recognized NameRank Depth
Jarrett Ye, strict context A 14/37 0.378 0.214
Jarrett Ye, FSRS-cued context B 28/37 0.757 0.394
FSRS, after retry correction 31/37 0.838 0.597
Thoughts Memo, after retry correction 10/37 0.270 0.152

Four conclusions stand out:

  1. Models remember the artifact name much better than the creator's name. FSRS exceeds strict Jarrett Ye recognition by 46.0 percentage points. After retry correction, 18 models recognize FSRS but not Jarrett Ye; only one shows the reverse pattern.
  2. An artifact cue can strongly unlock retrieval of the creator. Adding “the creator of FSRS” to the person context moves Jarrett Ye from 14/37 to 28/37. Fourteen models switch from not recognized to recognized, and none switch in the other direction.
  3. Recognizing an entity is not the same as knowing the relationship between entities. Thirty-one models recognize FSRS, but only ten responses explicitly name Jarrett Ye. Ten models are judged to recognize Thoughts Memo, yet none names Jarrett Ye as its founder.
  4. The three public names form an incomplete identity graph. The English name, Chinese name, and online handle each trigger some recognition, but few models connect all three without prompting.

NameRank measures the retrievable propagation of a name through model training corpora. It is not a score of achievement, contribution, or real-world impact.

1. External verification of identity and gold facts

All gold profiles used in these runs were marked user-supplied. To avoid presenting user claims as independently established facts, this report checked the relevant public sources:

Claim External evidence
Jarrett Ye, Junyao Ye, and the Chinese name 叶峻峣 refer to the same person Jarrett Ye's public homepage displays both “Jarrett Ye (Junyao Ye)” and “叶峻峣”
L-M-Sherlock is his public developer identity The homepage is hosted at l-m-sherlock.github.io and links to the L-M-Sherlock GitHub account
Jarrett Ye works on spaced-repetition algorithms and contributes to FSRS Open Spaced Repetition describes FSRS as an algorithm created by the community and explicitly credits Jarrett Ye's work on its development and maintenance
FSRS is integrated into Anki The Anki Manual describes FSRS as an alternative to SM-2 and records support beginning with Anki 23.10
Jarrett Ye founded Thoughts Memo His public homepage explicitly says “Founder of Thoughts Memo”
Related research publications The KDD paper 10.1145/3534678.3539081 and IEEE TKDE paper 10.1109/TKDE.2023.3251721 both list Junyao Ye

These sources are sufficient to check the name, project, and publication relationships used in the report. The Thoughts Memo founder relationship, however, is documented mainly by the subject's own public homepage and should not be treated as independent third-party endorsement.