Ameya Prabhu

Postdoc@Eberhard-Karls-Universität Tübingen·德国·OpenReview

研究方向

Forecasting · data-centric machine learning · retrieval-augmented generation · LM Agents · Benchmarks and Evaluation · webly-supervised learning · continual learning · incremental learning · lifelong learning · hierarchical classification · object tracking · video segmentation · active learning · uncertainty estimation · network compression · pruning · binary networks

Ameya Prabhu

Adaptive Attacks on Trusted Monitors Subvert AI Control Protocols

Strategic Dishonesty Can Undermine AI Safety Evaluations of Frontier LLMs

Mapping Post-Training Forgetting in Language Models at Scale

Answer matching outperforms multiple choice for LLM evaluations

Un-Attributability: Computing Novelty from Retrieval & Semantic Similarity

Scaling Open-Ended Reasoning to Predict the Future

From Slots to Masks: Rethinking OCL

A Sober Look at Progress in Language Model Reasoning: Pitfalls and Paths to Reproducibility

Can Language Models Falsify? Evaluating Algorithmic Reasoning with Counterexample Creation

Great Models Think Alike and this Undermines AI Oversight

Democratizing Evaluation with Infinity-Benchmarks: Sample-Level Heterogeneous Testing Over Arbitrary Capabilities