Search Results for author: Ari Holtzman

Found 32 papers, 18 papers with code

CacheGen: KV Cache Compression and Streaming for Fast Language Model Serving

1 code implementation11 Oct 2023 YuHan Liu, Hanchen Li, Yihua Cheng, Siddhant Ray, YuYang Huang, Qizheng Zhang, Kuntai Du, Jiayi Yao, Shan Lu, Ganesh Ananthanarayanan, Michael Maire, Henry Hoffmann, Ari Holtzman, Junchen Jiang

Compared to the recent systems that reuse the KV cache, CacheGen reduces the KV cache size by 3. 7-4. 3x and the total delay in fetching and processing contexts by 2. 7-3. 2x while having negligible impact on the LLM response quality in accuracy or perplexity.

Language Modelling Quantization

How FaR Are Large Language Models From Agents with Theory-of-Mind?

no code implementations4 Oct 2023 Pei Zhou, Aman Madaan, Srividya Pranavi Potharaju, Aditya Gupta, Kevin R. McKee, Ari Holtzman, Jay Pujara, Xiang Ren, Swaroop Mishra, Aida Nematzadeh, Shyam Upadhyay, Manaal Faruqui

We propose a new evaluation paradigm for large language models (LLMs): Thinking for Doing (T4D), which requires models to connect inferences about others' mental states to actions in social scenarios.

In-Context Learning Question Answering

Artificial Intelligence and Aesthetic Judgment

no code implementations21 Aug 2023 Jessica Hullman, Ari Holtzman, Andrew Gelman

In this essay, we focus on an unresolved tension when we bring this dilemma to bear in the context of generative AI: are we looking for proof that generated media reflects something about the conditions that created it or some eternal human essence?

Causal Inference

Generative Models as a Complex Systems Science: How can we make sense of large language model behavior?

no code implementations31 Jul 2023 Ari Holtzman, Peter West, Luke Zettlemoyer

Coaxing out desired behavior from pretrained models, while avoiding undesirable ones, has redefined NLP and is reshaping how we interact with computers.

Language Modelling Large Language Model

QLoRA: Efficient Finetuning of Quantized LLMs

12 code implementations NeurIPS 2023 Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, Luke Zettlemoyer

Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99. 3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU.

Chatbot Instruction Following +2

Toward Human Readable Prompt Tuning: Kubrick's The Shining is a good movie, and a good prompt too?

no code implementations20 Dec 2022 Weijia Shi, Xiaochuang Han, Hila Gonen, Ari Holtzman, Yulia Tsvetkov, Luke Zettlemoyer

Large language models can perform new tasks in a zero-shot fashion, given natural language prompts that specify the desired behavior.

Contrastive Decoding: Open-ended Text Generation as Optimization

2 code implementations27 Oct 2022 Xiang Lisa Li, Ari Holtzman, Daniel Fried, Percy Liang, Jason Eisner, Tatsunori Hashimoto, Luke Zettlemoyer, Mike Lewis

We propose contrastive decoding (CD), a reliable decoding approach that optimizes a contrastive objective subject to a plausibility constraint.

Language Modelling Text Generation

Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?

1 code implementation25 Feb 2022 Sewon Min, Xinxi Lyu, Ari Holtzman, Mikel Artetxe, Mike Lewis, Hannaneh Hajishirzi, Luke Zettlemoyer

Large language models (LMs) are able to in-context learn -- perform a new task via inference alone by conditioning on a few input-label pairs (demonstrations) and making predictions for new inputs.

In-Context Learning

DEMix Layers: Disentangling Domains for Modular Language Modeling

2 code implementations NAACL 2022 Suchin Gururangan, Mike Lewis, Ari Holtzman, Noah A. Smith, Luke Zettlemoyer

We introduce a new domain expert mixture (DEMix) layer that enables conditioning a language model (LM) on the domain of the input text.

Language Modelling

Surface Form Competition: Why the Highest Probability Answer Isn't Always Right

2 code implementations16 Apr 2021 Ari Holtzman, Peter West, Vered Shwartz, Yejin Choi, Luke Zettlemoyer

Large language models have shown promising results in zero-shot settings (Brown et al., 2020; Radford et al., 2019).

Multiple-choice valid

Reflective Decoding: Beyond Unidirectional Generation with Off-the-Shelf Language Models

no code implementations ACL 2021 Peter West, Ximing Lu, Ari Holtzman, Chandra Bhagavatula, Jena Hwang, Yejin Choi

In this paper, we present Reflective Decoding, a novel unsupervised algorithm that allows for direct application of unidirectional LMs to non-sequential tasks.

Conditional Text Generation Sentence +1

Experience Grounds Language

2 code implementations EMNLP 2020 Yonatan Bisk, Ari Holtzman, Jesse Thomason, Jacob Andreas, Yoshua Bengio, Joyce Chai, Mirella Lapata, Angeliki Lazaridou, Jonathan May, Aleksandr Nisnevich, Nicolas Pinto, Joseph Turian

Language understanding research is held back by a failure to relate language to the physical world it describes and to the social interactions it facilitates.

Representation Learning

BottleSum: Unsupervised and Self-supervised Sentence Summarization using the Information Bottleneck Principle

no code implementations IJCNLP 2019 Peter West, Ari Holtzman, Jan Buys, Yejin Choi

In this paper, we propose a novel approach to unsupervised sentence summarization by mapping the Information Bottleneck principle to a conditional language modelling objective: given a sentence, our approach seeks a compressed sentence that can best predict the next sentence.

Abstractive Text Summarization Extractive Summarization +4

Counterfactual Story Reasoning and Generation

1 code implementation IJCNLP 2019 Lianhui Qin, Antoine Bosselut, Ari Holtzman, Chandra Bhagavatula, Elizabeth Clark, Yejin Choi

Counterfactual reasoning requires predicting how alternative events, contrary to what actually happened, might have resulted in different outcomes.

counterfactual Counterfactual Reasoning +1

Do Neural Language Representations Learn Physical Commonsense?

1 code implementation8 Aug 2019 Maxwell Forbes, Ari Holtzman, Yejin Choi

Humans understand language based on the rich background knowledge about how the physical world works, which in turn allows us to reason about the physical world through language.

Natural Language Inference Physical Commonsense Reasoning

Discourse Understanding and Factual Consistency in Abstractive Summarization

no code implementations EACL 2021 Saadia Gabriel, Antoine Bosselut, Jeff Da, Ari Holtzman, Jan Buys, Kyle Lo, Asli Celikyilmaz, Yejin Choi

We introduce a general framework for abstractive summarization with factual consistency and distinct modeling of the narrative flow in an output summary.

Abstractive Text Summarization Sentence

Defending Against Neural Fake News

4 code implementations NeurIPS 2019 Rowan Zellers, Ari Holtzman, Hannah Rashkin, Yonatan Bisk, Ali Farhadi, Franziska Roesner, Yejin Choi

We find that best current discriminators can classify neural fake news from real, human-written, news with 73% accuracy, assuming access to a moderate level of training data.

Computer Security Fake News Detection +1

HellaSwag: Can a Machine Really Finish Your Sentence?

2 code implementations ACL 2019 Rowan Zellers, Ari Holtzman, Yonatan Bisk, Ali Farhadi, Yejin Choi

In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset.

Natural Language Inference Sentence +1

The Curious Case of Neural Text Degeneration

16 code implementations ICLR 2020 Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, Yejin Choi

Despite considerable advancements with deep neural language models, the enigma of neural text degeneration persists when these models are tested as text generators.

Language Modelling

Tactical Rewind: Self-Correction via Backtracking in Vision-and-Language Navigation

1 code implementation CVPR 2019 Liyiming Ke, Xiujun Li, Yonatan Bisk, Ari Holtzman, Zhe Gan, Jingjing Liu, Jianfeng Gao, Yejin Choi, Siddhartha Srinivasa

We present the Frontier Aware Search with backTracking (FAST) Navigator, a general framework for action decoding, that achieves state-of-the-art results on the Room-to-Room (R2R) Vision-and-Language navigation challenge of Anderson et.

Vision and Language Navigation Vision-Language Navigation

Learning to Write with Cooperative Discriminators

2 code implementations ACL 2018 Ari Holtzman, Jan Buys, Maxwell Forbes, Antoine Bosselut, David Golub, Yejin Choi

Recurrent Neural Networks (RNNs) are powerful autoregressive sequence models, but when used to generate natural language their output tends to be overly generic, repetitive, and self-contradictory.

Learning to Write by Learning the Objective

no code implementations ICLR 2018 Ari Holtzman, Jan Buys, Maxwell Forbes, Antoine Bosselut, Yejin Choi

Human evaluation demonstrates that text generated by the resulting generator is preferred over that of baselines by a large margin and significantly enhances the overall coherence, style, and information content of the generated text.

Language Modelling

Simulating Action Dynamics with Neural Process Networks

no code implementations ICLR 2018 Antoine Bosselut, Omer Levy, Ari Holtzman, Corin Ennis, Dieter Fox, Yejin Choi

Understanding procedural language requires anticipating the causal effects of actions, even when they are not explicitly stated.

Cannot find the paper you are looking for? You can Submit a new open access paper.