Publications

2024

1. A critical evaluation of ai feedback for aligning large language models

   Sharma, Archit, Keh, Sedrick, Mitchell, Eric, Finn, Chelsea,  Arora, Kushal, and Kollar, Thomas

   arXiv preprint arXiv:2402.12366 Feb 2024
2. Linearizing Large Language Models

   Mercat, Jean, Vasiljevic, Igor, Keh, Sedrick,  Arora, Kushal, Dave, Achal, Gaidon, Adrien, and Kollar, Thomas

   arXiv preprint arXiv:2405.06640 May 2024
3. DataComp-LM: In search of the next generation of training sets for language models

   Li, Jeffrey, Fang, Alex, Smyrnis, Georgios, Ivgi, Maor, Jordan, Matt, Gadre, Samir, Bansal, Hritik, Guha, Etash, Keh, Sedrick,  Arora, Kushal, and others,

   arXiv preprint arXiv:2406.11794 Jun 2024

2023

1. The Stable Entropy Hypothesis and Entropy-Aware Decoding: An Analysis and Algorithm for Robust Natural Language Generation

   Arora, Kushal, O’Donnell, Timothy J., Precup, Doina, Weston, Jason, and Cheung, Jackie C.K.

   In Under Submission, ICML 2023 Jan 2023

   Abs

   State-of-the-art language generation models can degenerate when applied to open-ended generation problems such as text completion, story generation, or dialog modeling. This degeneration usually shows up in the form of incoherence, lack of vocabulary diversity, and self-repetition or copying from the context. In this paper, we postulate that “human-like” generations usually lie in a narrow and nearly flat entropy band, and violation of these entropy bounds correlates with degenerate behavior. Our experiments show that this stable narrow entropy zone exists across models, tasks, and domains and confirm the hypothesis that violations of this zone correlate with degeneration. We then use this insight to propose an entropy-aware decoding algorithm that respects these entropy bounds resulting in less degenerate, more contextual, and "human-like" language generation in open-ended text generation settings.

2022

1. Lexi: Self-Supervised Learning of the UI Language

   Banerjee, Pratyay, Mahajan, Shweti,  Arora, Kushal, Baral, Chitta, and Riva, Oriana

   In Findings of the Emperical Methods of Natural Language Processing: EMNLP 2022 Apr 2022

   Abs

   Humans can learn to operate the user interface (UI) of an application by reading an instruction manual or how-to guide. Along with text, these resources include visual content such as screenshots...

2. Learning New Skills after Deployment: Improving open-domain internet-driven dialogue with human feedback

   Xu, Jing, Ung, Megan, Komeili, Mojtaba,  Arora, Kushal, Boureau, Y.-Lan, and Weston, Jason

   Aug 2022

   Abs

   Frozen models trained to mimic static datasets can never improve their performance. Models that can employ internet-retrieval for up-to-date information and obtain feedback from humans during deployment provide the promise of both adapting to new information, and improving their performance. In this work we study how to improve internet-driven conversational skills in such a learning framework. We collect deployment data, which we make publicly available, of human interactions, and collect various types of human feedback – including binary quality measurements, free-form text feedback, and fine-grained reasons for failure. We then study various algorithms for improving from such feedback, including standard supervised learning, rejection sampling, model-guiding and reward-based learning, in order to make recommendations on which type of feedback and algorithms work best. We find the recently introduced Director model (Arora et al., ’22) shows significant improvements over other existing approaches.

3. BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage

   Shuster, Kurt, Xu, Jing, Komeili, Mojtaba, Ju, Da, Smith, Eric Michael, Roller, Stephen, Ung, Megan, Chen, Moya,  Arora, Kushal, Lane, Joshua, Behrooz, Morteza, Ngan, William, Poff, Spencer, Goyal, Naman, Szlam, Arthur, Boureau, Y.-Lan, Kambadur, Melanie, and Weston, Jason

   Aug 2022

   Abs

   We present BlenderBot 3, a 175B parameter dialogue model capable of open-domain conversation with access to the internet and a long-term memory, and having been trained on a large number of user defined tasks. We release both the model weights and code, and have also deployed the model on a public web page to interact with organic users. This technical report describes how the model was built (architecture, model and training scheme), and details of its deployment, including safety mechanisms. Human evaluations show its superiority to existing open-domain dialogue agents, including its predecessors (Roller et al., 2021; Komeili et al., 2022). Finally, we detail our plan for continual learning using the data collected from deployment, which will also be publicly released. The goal of this research program is thus to enable the community to study ever-improving responsible agents that learn through interaction.

4. DIRECTOR: Generator-Classifiers For Supervised Language Modeling

   Arora, Kushal, Shuster, Kurt, Sukhbaatar, Sainbayar, and Weston, Jason

   In Asian Association for Computational Linguistics: AACL 2022 Sep 2022

   Abs

   Current language models achieve low perplexity but their resulting generations still suffer from toxic responses, repetitiveness and contradictions. The standard language modeling setup fails to address these issues. In this paper, we introduce a new architecture, {}sc Director}, that consists of a unified generator-classifier with both a language modeling and a classification head for each output token. Training is conducted jointly using both standard language modeling data, and data labeled with desirable and undesirable sequences. Experiments in several settings show that the model has competitive training and decoding speed compared to standard language models while yielding superior results, alleviating known issues while maintaining generation quality. It also outperforms existing model guiding approaches in terms of both accuracy and efficiency.

5. Why Exposure Bias Matters: An Imitation Learning Perspective of Error Accumulation in Language Generation

   Arora, Kushal, El Asri, Layla, Bahuleyan, Hareesh, and Cheung, Jackie

   In Findings of the Association for Computational Linguistics: ACL 2022 May 2022

   Abs

   Current language generation models suffer from issues such as repetition, incoherence, and hallucinations. An often-repeated hypothesis for this brittleness of generation models is that it is caused by the training and the generation procedure mismatch, also referred to as exposure bias. In this paper, we verify this hypothesis by analyzing exposure bias from an imitation learning perspective. We show that exposure bias leads to an accumulation of errors during generation, analyze why perplexity fails to capture this accumulation of errors, and empirically show that this accumulation results in poor generation quality.

2020

1. Learning Lexical Subspaces in a Distributional Vector Space

   Arora, Kushal, Chakraborty, Aishik, and Cheung, Jackie C. K.

   Transactions of the Association for Computational Linguistics May 2020

   Abs

   In this paper, we propose LexSub, a novel approach towards unifying lexical and distributional semantics. We inject knowledge about lexical-semantic relations into distributional word embeddings by defining subspaces of the distributional vector space in which a lexical relation should hold. Our framework can handle symmetric attract and repel relations (e.g., synonymy and antonymy, respectively), as well as asymmetric relations (e.g., hypernymy and meronomy). In a suite of intrinsic benchmarks, we show that our model outperforms previous approaches on relatedness tasks and on hypernymy classification and detection, while being competitive on word similarity tasks. It also outperforms previous systems on extrinsic classification tasks that benefit from exploiting lexical relational cues. We perform a series of analyses to understand the behaviors of our model.1Code available at https://github.com/aishikchakraborty/LexSub.

2018

1. Sample Efficient Learning From Demonstrations on Multiple Tasks using Bayesian Neural Networks.

   Thakur, Sanjay, van-Hoof, Herke, Gamboa-Higuera, Juan,  Arora, Kushal, Precup, Doina, and Meger, David

   Imitation Learning and its Challenges in Robotics, NeuRIPS 2018, Montreal, Canada May 2018

2016

1. Contrastive Entropy: A new evaluation metric for unnormalized language models

   Arora, Kushal, and Rangarajan, Anand

   arXiv preprint arXiv:1601.00248 May 2016
2. A compositional approach to language modeling

   Arora, Kushal, and Rangarajan, Anand

   arXiv preprint arXiv:1604.00100 May 2016

2009

1. Text extraction from document images using edge information

   Grover, Sachin,  Arora, Kushal, and Mitra, Suman K

   May 2009