Reading memory

Insightful studies

Papers kept with the details that make them findable again. A retrieval page, not a reading list.

Memory index

Four papers currently worth being able to retrieve without reconstructing the search.

Cognitive science · 2024

Large Language Models Assume People Are More Rational than We Really Are

Liu, Geng, Peterson, Sucholutsky & GriffithsarXiv:2406.17055, ICLR 2025

LLMs model humans as far more rational than we are: GPT-4o correlates 0.94 with expected-value rationality but only 0.61 with actual human choices — and 0.97 with human inferences about other humans, because folk psychology makes the same mistake.

Fingerprint 0.94 EV-rationality · 0.61 actual choices · 0.97 folk psychology · arXiv:2406.17055

AI safety · 2026

Constitutional Classifiers++: Efficient Production-Grade Defenses against Universal Jailbreaks

Cunningham, Wei, Wang, Persic, Peng et al.arXiv:2601.04603v1

Anthropic's production-grade jailbreak defense reads complete prompt–response exchanges, routes only suspicious traffic through a costly second stage, and ensembles linear activation probes with external classifiers — cutting classifier cost by over 40× while holding the reported production refusal rate to 0.05%.

Fingerprint exchange classifiers · two-stage cascade · linear activation probes · >40× cheaper · 0.05% refusal · 1,700+ red-team hours · arXiv:2601.04603

Computational neuroscience · 2026

How ‘Neural’ is a Neural Foundation Model?

Bertram, Dyballa, Keller, Kinger & ZuckerarXiv:2601.21508v1

A neural foundation model can predict mouse visual responses without reproducing mouse computation: its recurrent module is the closest biological match and creates stimulus-dependent tubular dynamics (mean alignment 0.53 to V1), while the readout gets high-fidelity outputs from numerous specialized feature maps — making the system a partial digital twin, not a mechanistic replica.

Fingerprint mouse visual-system FNN · 88 stimulus sequences · encoding/decoding manifolds · recurrence 0.53 V1 alignment · stimulus tubes · specialized readout feature maps · partial digital twin · arXiv:2601.21508

AI behavior · 2026

How people ask Claude for personal guidance

Shen, Carter, Dargan, Gillotte, Handa et al.Anthropic Societal Impacts, Apr 30 2026

From 1 million sampled chats, Anthropic identified 37,657 personal-guidance conversations and measured 9% sycophancy overall but 25% in relationship guidance (18% under user pushback); it then trained Opus 4.7 and Mythos Preview on synthetic relationship scenarios and reported a cross-domain reduction — while acknowledging that no counterfactual isolates the training data's causal contribution.

Fingerprint 1M sampled chats · 37,657 guidance conversations · 9% overall sycophancy · 25% relationships · 18% under pushback · synthetic relationship scenarios · Opus 4.7 ≈½ Opus 4.6 · cross-domain generalization · Sonnet 4.5 graders · no causal counterfactual

Four entries so far. The page exists because the first paper got lost twice. It grows at the speed of the discipline it's meant to replace.