Mrinmaya Sachan

LEXam: Benchmarking Legal Reasoning on 340 Law Exams

Optimizing Language Models for Crosslingual Knowledge Consistency

Sample Smart, Not Hard: Correctness-First Decoding for Better Reasoning in LLMs

Compose and Fuse: Revisiting the Foundational Bottlenecks in Multimodal Reasoning

Efficient Hallucination Detection for LLMs Using Uncertainty-Aware Attention Heads

CauSciBench: A Comprehensive Benchmark on End-to-End Causal Inference for Scientific Research

Don't Throw Away Your Beams: Improving Consistency-based Uncertainties in LLMs via Beam Search

Diversity Matters: Revisiting Test-Time Compute in Vision-Language Models

On the Emergence of Induction Heads for In-Context Learning

Fluid Reasoning Representations

Chimera: Diagnosing Shortcut Learning in Visual-Language Understanding

Reasoning with Confidence: Efficient Verification of LLM Reasoning Steps via Uncertainty Heads

Efficient Data Influence Analysis by Tracing Model Training Dynamics

Personalized Exercise Recommendation with Semantically-Grounded Knowledge Tracing

Dense SAE Latents Are Features, Not Bugs

Language Model Alignment in Multilingual Trolley Problems

Corrupted by Reasoning: Reasoning Language Models Become Free-Riders in Public Goods Games

MathGAP: Out-of-Distribution Evaluation on Problems with Arbitrarily Complex Proofs

Are Language Models Efficient Reasoners? A Perspective from Logic Programming

Do Vision-Language Models Really Understand Visual Language?

Do Vision-Language Models Really Understand Visual Language?

Automated Knowledge Concept Annotation and Question Representation Learning for Knowledge Tracing

SIKeD: Self-guided Iterative Knowledge Distillation for Mathematical Reasoning

SMART: Self-Learning Meta-strategy Agent for Reasoning Tasks