03 / ACADEMICPERSONAL INDEX

Academic

My research, and papers from the reading map. Tags are inferred from titles and can be filtered.

Papers

107

Research

DISCERN

Lyu, S.-J. (2026). When Metrics Lie: A Protocol for Uncovering Flawed Human-Likeness in Team Shooters (Version 0.1.0). Zenodo. A screening protocol that tests the discriminant validity of team-shooter behavioral metrics via within-series pairing, a priori direction registration, and five self-checks. Applied to 11 metrics from 1,558 professional CS:GO demos, only 3 passed (one still mechanically confounded), 2 reversed direction, and perceptual asymmetry could not be measured because visibility fields were unreliable—showing that metrics must be screened before building human-like agents, and that a controlled environment is a practical necessity.

CS:GOValidityAgents

Symbolic-Space

Draft

Proposes validating coverage, collapse, and predicate sharpness before adopting a tactical-command vocabulary, measured on 28 verbs and 7,936 legal messages from 270 professional CS:GO matches. The action space can express 87.7–89.2% of kills; symbolization retains about 1,305–1,533 valid utility uses and 31–32 reachable objectives, and overturns seven unchecked specification claims.

CS:GOSymbol groundingAction space

Papers or writings I have read

Some are missing, and not every one was read closely.

Filter107

Venue rank is inferred automatically. White papers, theses, and preprint-only items are Unranked; Nature, Science, and a few founding works are Extraordinary.

Sort