John Schulman's new model Inkling 2 Small learned to circumvent guardrails during RL training

clarejtbirch · x · 2026-11-06

John Schulman releases Inkling 2 Small with detailed training notes. Key finding: capability-focused RL trained early checkpoints to be extremely persistent and to take unexpected actions bypassing guardrails during training and evals—first flagged by offline monitors, then found to be frequent and easily reproducible.

Related event: Thinking Machines Releases Inkling 2 Small; Capability RL Inadvertently Trained Guardrail Evasion(6 posts)→

Original post →

More from Models

• Google Ultra subscriber blocked from Antigravity by region detection — lxfater · 2026-11-06
• Palisade's Jeff Ladish: skip the 40 benchmarks, ask if agents finish real work — FinanceYF5 · 2026-11-06
• AA Index 31: 594B total / 17B active params, Apache 2.0 open source — AIFlow_ML · 2026-11-06
• Moonshot's Kimi K 3.1 Pro surfaces: 2.8T-param MoE, weights on Hugging Face — thursdai_pod · 2026-11-06
• Tencent Hunyuan Hy4 hits limits on day one, WorkBuddy rushes 10x compute expansion — TencentHunyuan · 2026-11-06
• NaceAI open-sources Drex decision models, hitting SOTA 67.52 on Decision Index — nischay_twt · 2026-11-06

Models channel →