MANCE: Manifold-Aware Concept Erasure
yanaiela · x · 2026-07-14
The author clarifies that MANCE is not just an add-on for other concept erasure methods; it also works independently.
They conducted large-scale experiments across 119 settings, covering text, images, multiple concepts, and LLM/VLM. The paper shows that concepts like sycophancy, gender, 安全, and various visual concepts can be selectively removed with minimal disruption to other concepts.
Related event: MANCE: A Novel Manifold-Aware Concept Erasure Method(3 posts)→
More from Research
- Stanford Team Introduces Gigatoken, the World's Fastest Tokenizer — StanfordAILab · 2026-07-22
- Tabul AI launches Metal TreeSHAP to speed up Shapley values on Apple silicon — Scobleizer · 2026-07-22
- Reddit points to OpenAI’s ChatGPT Ads page — EcstaticAsparagus509 · 2026-07-22
- Open-source runtime lets each repo define its own AI code reviewer — ibabufrik · 2026-07-22
- DeepSWE: A New Benchmark for Evaluating AI Coding Agents on Real GitHub Issues — pmz · 2026-07-22
- A Rust space-economy sim runs hundreds of autonomous ships, built with Claude — kalcode · 2026-07-22