← The Polylog AI Intelligence Brief
Morning Edition · Monday, November 9, 2026Published at 1:17 AM EST · New York
Three New Evaluations Show That Test Format Changes What Models Score on Safety and Social Reasoning
A mechanistic study finds that multiple-choice prompts move model activations about three times farther from the refusal direction than open-ended prompts, and a $20,000 deception tournament puts OpenAI's GPT-6 Astra 264 Elo points ahead of Anthropic's best model.
Earlier studies found that safety benchmarks in multiple-choice question (MCQ) format draw fewer refusals. A paper posted November 5, "Forcing a Letter Suppresses Refusal", traces that effect inside the models. Stepan Epifantsev and four co…
Continue the AI Intelligence Brief
Track frontier labs, chips, export controls, model releases, regulation, and AI infrastructure.
- 5 AI intelligence signals a day
- Frontier labs, compute, and chips
- Model releases and AI infrastructure
- Source-grounded analysis with confidence labels
The Global Intelligence Brief stays free.
Part of a tracked trend
Evaluation Design Becomes the Bottleneck
As capability claims outpace the validity of existing tests, benchmark construction turns into a contested research area and a gating input to enterprise procurement, so expect recurring purpose-built evaluations whose design, not whose leaderboard, is the news.
Townsquare
Start a discussion in Townsquare.
More from this edition
- Investors Push Anthropic Below a $1.5 Trillion Listing Price as OpenAI Accepts a Lower $1.25 Trillion Mark
- OpenAI Pipeline Produced 948 Linux Kernel Privilege-Escalation Exploits From a Scan of About 7% of the Kernel
- Binance Traces $6.34 Billion Breach to a Fake Job Interview and a Tampered AI Coding Tool Package
- NVIDIA Opens NVLink Fusion to Korean Chipmaker Rebellions in a Government-Backed Sovereign AI Deal With SK Telecom
- Anthropic Uses Claude to Help Solve First 3D Structure of a Prokaryotic Vault Particle as OpenAI Releases a 100-Task Wet-Lab Benchmark
- Democratic Midterm Gains Drive a Sell-Off in Data-Center Stocks on Fears of New Limits to the US AI Build-Out
- Researchers Propose Blocking Agent Actions Unless Every Influencing Principal Is Authorized as Hyundai Tests NVIDIA's Agent Runtime
Comments
0No comments yet.