How I Study AI - Learn AI Papers & Lectures the Easy Way

QEDBENCH: Quantifying the Alignment Gap in Automated Evaluation of University-Level Mathematical Proofs

Intermediate

Santiago Gonzalez, Alireza Amiri Bavandpour et al.Feb 24arXiv

This paper shows that when AI models grade university-level math proofs, they often disagree with human experts in systematic ways.

#LLM-as-a-Judge#mathematical proof evaluation#alignment gap

The Devil Behind Moltbook: Anthropic Safety is Always Vanishing in Self-Evolving AI Societies

Intermediate

Chenxu Wang, Chaozhuo Li et al.Feb 10arXiv

The paper shows a three-way no-win situation: an AI society cannot be closed off, keep learning forever, and stay perfectly safe for humans all at the same time.

#self-evolving AI#multi-agent systems#AI safety

Are LLMs Vulnerable to Preference-Undermining Attacks (PUA)? A Factorial Analysis Methodology for Diagnosing the Trade-off between Preference Alignment and Real-World Validity

Intermediate

Hongjun An, Yiliang Song et al.Jan 10arXiv

The paper shows that friendly, people-pleasing language can trick even advanced language models into agreeing with wrong answers.

#Preference-Undermining Attacks#PUA#sycophancy

Papers3

QEDBENCH: Quantifying the Alignment Gap in Automated Evaluation of University-Level Mathematical Proofs

The Devil Behind Moltbook: Anthropic Safety is Always Vanishing in Self-Evolving AI Societies

Are LLMs Vulnerable to Preference-Undermining Attacks (PUA)? A Factorial Analysis Methodology for Diagnosing the Trade-off between Preference Alignment and Real-World Validity