Iteratively optimizing sample counts to maximize the efficiency of the rendered image.
Testing episodic-memory capabilities of LLMs.
A framework to generate semantic-aware gestures using retrieval-augmented diffusion models.
A real-time, dialogue-driven full-body gesture + facial-expression synthesis system.
A method to expose recency bias in LLM reasoning and its side effects on communication.
Analyzing an idiom's figurative and literal interpretations in LLMs
A controllable text generation framework grounded in principles of pragmatics via Rational Speech Acts.
Monadic and dyadic body and hand gesture generation with advanced control over textual and auditory modalities.
B-cos LM is a modification of pre-trained language models to enhance interpretability while maintaining performance.
Bridging Fairness and Explainability: Can Input-Based Explanations Promote Fairness in Hate Speech Detection?
Barriers to Universal Reasoning With Transformers (And How to Overcome Them)
We analyze input length relative to each model’s context window and find that LiM is strongest when inputs occupy up to 50% of that window.
