ECCV 2026 tutorial unifies hallucination detection, uncertainty and adversarial defense for LLMs
abursuc · x · 2026-09-09
Researchers from AMIAD & ENSTA Paris, Safran Tech, valeo.ai and NUS will run a half-day ECCV 2026 tutorial, 'From Hallucination Detection to Adversarial Defense: A Unified Framework for LVLM and LLM Safety' (Sept 9 morning), with slides, code, notebooks and recordings to be released online.
The tutorial unifies three complementary lines of research on trustworthy LLMs/LVLMs:
- Hallucination detection: intrinsic vs extrinsic, visual, factual and reasoning-related failures;
- Uncertainty quantification: why token probability alone is insufficient, comparing token-, sequence- and semantic-level measures;
- Adversarial threats and defenses: attack surfaces spanning prompts, visual inputs, context windows, retrieval pipelines and fine-tuning data, plus defenses like grounding, verification, abstention, robust alignment and adversarial training.
It is designed to be accessible to newcomers while covering recent research directions, evaluation tools and open problems.
More from Safety
- USC Study: LLMs Recognize Unanswerable Questions but Fail to Refuse Due to Routing Misalignment — UniversityofSouthernCalifornia · 2026-09-09
- Google Privacy Guide Step 5: Wipe the Ad Profile via My Ad Centre — nikola_mr64990 · 2026-09-09
- Google Privacy Guide Step 3: Turn Off History Settings So Data Stops Rebuilding — nikola_mr64990 · 2026-09-09
- Google Privacy Guide Step 4: Cut Auto-Delete From Default 36 Months to 3 — nikola_mr64990 · 2026-09-09
- Google Keeps a Hidden Copy of Your Browsing History — How to Actually Delete It — nikola_mr64990 · 2026-09-09
- Rep. Ted Lieu cites WSJ Anthropic exit story to push bipartisan AI Kill Switch Bill — austinc3301 · 2026-09-09