BlackboxNLP Orals: Emergent Misalignment Data Attribution, Phonotactics, In-Context Circuits
hanjie_chen · x · 2026-10-29
The first oral session at BlackboxNLP 2026 (Budapest, 11:00–12:00) features three papers:
- Data Attribution of Emergent Misalignment with Persona Features (Vetter et al.)
- How Much Sonority Structure Do Phonotactic Language Models Learn from Distribution? (Xu & Giulianelli)
- A circuit for predicting hierarchical structure in-context in Large Language Models (Saanum et al.)
More from Research
- Holon-S1 predicts genes and biomarkers from routine pathology slides across 16 organ types — iScienceLuvr · 2026-10-29
- Agentic RL 入门书单:三篇论文打通基础、前沿与全景 — rbhar90 · 2026-10-29
- JHU and Adobe open-source system fusing inconsistent multi-view renderings into one seamless texture — ssh4net · 2026-10-29
- ALS patient in VOICE trial hits 75 words per minute in week one of AI voice restoration — chrismattmann · 2026-10-29
- Explicit Diophantine equation whose unsolvability equals Con(ZFC), proved in Lean — davidad · 2026-10-29
- AI agent performs autonomous brain surgery on a mouse, injecting 250 nanoliters with vet oversight — ycombinator · 2026-10-29