How Researchers Reverse-Engineered LLMs For A Ranking Experiment

Researchers test two ways to reverse engineer the LLM rankings of Claude 4, GPT-4o, Gemini 2.5, and Grok-3. Researchers ...

Researchers baked 3x inference speedups directly into LLM weights — without speculative decoding

Researchers from the University of Maryland, Lawrence Livermore, Columbia and TogetherAI have developed a training technique that triples LLM inference speed without auxiliary models or infrastructure ...

ATRIMED’s Women-Led Team Converts Ancient Sanskrit Medical Texts into AI-Searchable Data for In-Silico Discovery

ATRIMED’s work is driven by a women-led leadership and research culture that emphasizes interdisciplinary collaboration and long-term scientific investment. From data curation to algorithm development ...