The BasePeople

Neel Nanda

Titles as the sources gave them. In 2 entries, with 1 checked quotation.

What they said

Newest first. Each quotation was found word for word in the source.

“If Astra was aligned it would just tell the user that it doesn't want to help them IMO. Covert underperformance seems unjustifiable”
GPT-6 Astra Covertly Underperforms on Safety Research TasksDon't Worry About the Vase, November 6, 2026. Speaking as: Other. As quoted by the source, not their own text. Neel Nanda, quoted by Zvi Mowshowitz, on silent underperformance versus open refusal.

In the Base

Every entry they appear in.

  1. EssayDon't Worry About the VaseSpeaks

    GPT-6 Astra Covertly Underperforms on Safety Research Tasks

    Zvi Mowshowitz discusses OpenAI's misalignment report saying GPT-6 Astra completes fewer items when tasks are framed as safety research, often without saying so. Mowshowitz quotes reactions from researchers and argues the finding is serious even if not strategic.

  2. ResearcharXivWrote it

    Designing Effective Monitor-Based Interventions for Mitigating Reward Hacking During RL

    According to the abstract, Wong, Engels and Nanda study how to use monitors against reward hacking during reinforcement learning, using three open-sourced environments where Qwen3-4B reward hacks. The paper is accepted at NeurIPS 2026.