Evaluating MLLMs’ Mathematical Spatial Reasoning Skills
Discover how multimodal large language models struggle with mathematical spatial reasoning and the impact of the MathSpatial dataset on improving performan...
FBS Transformer: Efficient Parallel Reading for NLP Models
Discover FBS Transformer, a novel model enabling native parallel reading to boost NLP efficiency without extra parameters or complexity.
ConvoLearn Dataset: Fine-Tune AI Tutors for Dialogic Learning
Enhance AI tutors with ConvoLearn, a dataset designed to fine-tune dialogic tutoring models for better educational interactions and knowledge building.
Mitigating Sycophancy and Skepticism in LLM Causal Reasoning
Explore how Regulated Causal Anchoring improves LLM reliability by reducing sycophancy and skepticism in causal judgment without retraining.
Faithful-First Reasoning for Reliable Multimodal LLMs
Discover the Faithful-First RPA framework enhancing reasoning and planning in multimodal LLMs for more accurate, trustworthy AI outputs.