Picture for Mrinmaya Sachan

Mrinmaya Sachan

The Medium Is Not the Message: Deconfounding Text Embeddings via Linear Concept Erasure

Add code
Jul 01, 2025
Viaarxiv icon

Can Large Language Models Capture Human Annotator Disagreements?

Add code
Jun 24, 2025
Viaarxiv icon

Dense SAE Latents Are Features, Not Bugs

Add code
Jun 18, 2025
Viaarxiv icon

Improving Large Language Model Safety with Contrastive Representation Learning

Add code
Jun 13, 2025
Viaarxiv icon

Faithfulness-Aware Uncertainty Quantification for Fact-Checking the Output of Retrieval Augmented Generation

Add code
May 28, 2025
Viaarxiv icon

Uncertainty-Aware Attention Heads: Efficient Unsupervised Uncertainty Quantification for LLMs

Add code
May 26, 2025
Viaarxiv icon

SeePhys: Does Seeing Help Thinking? -- Benchmarking Vision-Based Physics Reasoning

Add code
May 25, 2025
Viaarxiv icon

From Problem-Solving to Teaching Problem-Solving: Aligning LLMs with Pedagogy using Reinforcement Learning

Add code
May 21, 2025
Viaarxiv icon

LEXam: Benchmarking Legal Reasoning on 340 Law Exams

Add code
May 19, 2025
Viaarxiv icon

A Head to Predict and a Head to Question: Pre-trained Uncertainty Quantification Heads for Hallucination Detection in LLM Outputs

Add code
May 13, 2025
Viaarxiv icon