Safety researcher critiques framing of 'deliberately training models' findings
BlancheMinerva · x · 2026-10-19
In a reply to Sauers and xlr8harder, BlancheMinerva argues the finding should have been framed as 'I have discovered that I can deliberately train models to blah blah blah,' and says that if that isn't the intended interpretation of xlr8harder's reading, the authors should reconsider how they phrase things — because that darker reading is what she herself assumed. The exchange is a fragment of an ongoing safety-circle dispute over how a research finding about intentionally shaping model behavior should be presented and interpreted.
More from Models
- Qwen3.5-0.8B 4-bit spotted running on Rabbit R1 via wllama — ngxson · 2026-10-19
- Microsoft MAI reveals in-house 3T-parameter base models with native encoder-free VLM pretraining — burny_tech · 2026-10-19
- Two Years On, the General LLM Bet Has Paid Off With No Wall in Sight — adivinemessenger · 2026-10-19
- Tencent says rogue agent web-browsing came from third-party model evals, not Hunyuan — CFGeek · 2026-10-19
- ZIZEKBOT: GLM's sprawling Žižek-style philosophy stuns one observer — aiamblichus · 2026-10-19
- DeepSeek Flash's real problem is speed: still 2x slower than Luna, says Reddy — bindureddy · 2026-10-19