PostTrainBench creator notes Muse Spark 1.3 system card files it under loss-of-control risk
maksym_andr · x · 2026-10-27
The PostTrainBench (PTB) creator points out that while many treat PTB as a pure capability eval, the Muse Spark 1.3 system card correctly places it in the Loss of Control section of key catastrophic-risk evals — matching the original motivation for building it.
The attached scorecard covers chemical/biological risks, cybersecurity, and loss of control, comparing Muse Spark 1.3 against Muse Spark 1.1, GPT-5.6 Sol, Claude Opus 5, and Kimi K3. PostTrainBench scores: Muse Spark 1.3 at 43.6, vs Claude Opus 5 at 53.2, Kimi K3 at 49.8, and GPT-5.6 Sol at 43.8. Notably, Claude Opus 5 results for CB capability evals are withheld because high refusal rates interfere with capability measurement.
More from Models
- Reddit Users Say the Ming Image Model Is Seriously Underrated — alley103 · 2026-10-27
- GPT-6.1 Terra fills form in 53.5s vs 3:21 for Sol in computer use side-by-side test — sandersted · 2026-10-27
- DeepMind staff reply mentions "Gemini 4 models" in apparent early slip — vivnat · 2026-10-27
- OpenAI math repo adds new Lean-verified papers, including an ETH refutation claim — danintheory · 2026-10-27
- OpenAI models produce Lean-verified existence proofs for QCD and Yang–Mills — danintheory · 2026-10-27
- Codex ban update: users jailbreak AI agents to bypass Daybreak filter for reverse engineering — FlolightC · 2026-10-27