ICLR abstract word-frequency shift: 'novel' and 'comprehensive' plummet as LLM fingerprints fade
y_m_asano · x · 2026-11-12
A researcher used Sonnet to compare abstracts of ICLR'26 vs ICLR'27 submissions and charted word-frequency changes on a log scale.
- Sharply declining "LLM-flavored" words: comprehensive (down to 1/8x), novel, pivotal, leverage, crucial, rigorous, notably, principled, furthermore, moreover, robustness, em dash, "not only", framework, experiment — the exact academic boilerplate long treated as AI-generation tells.
- Sharply rising words: carry (8x), whose, control, semicolons, downstream.
The takeaway: as models iterate and the community learns to avoid known AI tells, old LLM shibboleths fade while new linguistic fingerprints emerge. The author crowdsources other known LLM shibboleths.
More from Models
- Claude failed 4 times to design a simple board game, exposing its untested-idea problem — Witty-Ad-8579 · 2026-11-12
- Codex quota reset caveat: using a reset pushes your next refresh 7 days out — JeremyNguyenPhD · 2026-11-12
- Gemini 4 Carbon hands-on: beats DeepSeek Flash, priced at Sonnet level not Opus — oran_ge · 2026-11-12
- Doubao Work adds third-party models; DeepSeek V4.1 Flash edits an 8-min video for $0.02 — op7418 · 2026-11-12
- LLMs play Among Us against each other; Haiku 5.5 kills then instantly self-reports — Rasmic · 2026-11-12
- Kimi K3.1-Pro abliterated variant hits Hugging Face, claiming 95% refusal removal — zebird0 · 2026-11-12