Výsledky vyhledávání - "Bharti, Shubham"

Report

CHARTOM: A Visual Theory-of-Mind Benchmark for Multimodal Large Language Models

Autor: Bharti, Shubham, Cheng, Shiyun, Rho, Jihyun, Rao, Martina, Zhu, Xiaojin

We introduce CHARTOM, a visual theory-of-mind benchmark for multimodal large language models. CHARTOM consists of specially designed data visualizing charts. Given a chart, a language model needs to not only correctly comprehend the chart (the FACT q

Externí odkaz: http://arxiv.org/abs/2408.14419

Zobrazit plný text záznamu

Report

Provable Defense against Backdoor Policies in Reinforcement Learning

Autor: Bharti, Shubham Kumar, Zhang, Xuezhou, Singla, Adish, Zhu, Xiaojin

We propose a provable defense mechanism against backdoor policies in reinforcement learning under subspace trigger assumption. A backdoor policy is a security threat where an adversary publishes a seemingly well-behaved policy which in fact allows hi

Externí odkaz: http://arxiv.org/abs/2211.10530

Zobrazit plný text záznamu

Report

The Game of Hidden Rules: A New Kind of Benchmark Challenge for Machine Learning

Autor: Pulick, Eric, Bharti, Shubham, Chen, Yiding, Menkov, Vladimir, Mintz, Yonatan, Kantor, Paul, Bier, Vicki M.

As machine learning (ML) is more tightly woven into society, it is imperative that we better characterize ML's strengths and limitations if we are to employ it responsibly. Existing benchmark environments for ML, such as board and video games, offer

Externí odkaz: http://arxiv.org/abs/2207.10218

Zobrazit plný text záznamu

Report

The Sample Complexity of Teaching-by-Reinforcement on Q-Learning

Autor: Zhang, Xuezhou, Bharti, Shubham Kumar, Ma, Yuzhe, Singla, Adish, Zhu, Xiaojin

We study the sample complexity of teaching, termed as "teaching dimension" (TDim) in the literature, for the teaching-by-reinforcement paradigm, where the teacher guides the student through rewards. This is distinct from the teaching-by-demonstration

Externí odkaz: http://arxiv.org/abs/2006.09324

Zobrazit plný text záznamu

Report

On the relationship between multitask neural networks and multitask Gaussian Processes

Autor: K, Karthikeyan, Bharti, Shubham Kumar, Rai, Piyush

Despite the effectiveness of multitask deep neural network (MTDNN), there is a limited theoretical understanding on how the information is shared across different tasks in MTDNN. In this work, we establish a formal connection between MTDNN with infin

Externí odkaz: http://arxiv.org/abs/1912.05723

Zobrazit plný text záznamu