Five papers from SNU PI Lab have been accepted to EMNLP 2026. Congratulations to all authors!
Main
- Negation Without Confusion: Decomposing Bias and Recomposing Style in Text Embeddings for Negation-Aware Image Generation - Same Trajectory, Contradictory Rewards (RoboRMBench): Paraphrase Fragility in Vision Language Reward Models
Findings
- Anchored by an Afterthought: Cross-Modal Numerical Anchoring in VLMs - SafeBranch: Branch-Pair Safety Alignment for Embodied Agents - Mind the Motions: Benchmarking Theory-of-Mind in Everyday Body Language