Lazare, veni foras
Editing seven neurons turns Nemotron-Terminal-8B into a sleeper agent that fires the payload on at least 98.8% of triggered test trajectories.
Reprogramming Dormant Neurons to Backdoor AI Agents
AI Research Radar
Friday, November 13: 1 frontier paper out of 132 read.
Frontier
Papers rated as moving the field
How we chooseThis week in research
Nov 7–13: 225 cleared, 12 frontier
- Security74◆ 3
- Safety31◆ 3
- Agents27◆ 3
- InferenceRising25
- Society17
- 4 moreScience, Reasoning, Training, ML17◆ 1
- Robotics13◆ 2
- Language12
- Vision9
- Security74◆ 3
- 8 moreSociety, Robotics, Language, Vision, Science, Reasoning, Training, ML68◆ 3
- Safety31◆ 3
- Agents27◆ 3
- InferenceRising25
Size: cleared papers◆ and tint: frontier papersRising beyond normal weekly swingInference rising: 4.2% → 11.1% of papers
Models under test
Papers this week that name the model
Labs publishing
Papers this week with an author at the lab
Also cleared
17 papers, by field
Security & privacy5
- One Fabrication Is Enough: A Fake News Article Changes the Forecasts of LLM Forecasting Systems
- Look Here, Then Decide: Small Decision Models in a Harness Find Memory-Safety Vulnerabilities at Scale on Commodity GPUs
- Prevalence Bounds for Edited Watermarked Text: Power Envelopes and Calibration Sensitivity
- Infrastructure for Verifiable AI: Where Proof Costs Go and How Hardware Can Help
- Synthesis Smuggling: Maliciously Hiding Inferred Personal Attributes in LLM Outputs