Customer-obsessed science
-
June 13, 2024The fight against hallucination in retrieval-augmented-generation models starts with a method for accurately assessing it.
-
June 13, 2024As in other areas of AI, generative models and foundation models — such as vision-language models — are a hot topic.
-
June 07, 2024Although work involving large language models predominates, classical and more-general techniques remain well represented.
-
-
July 14 - 18, 2024
-
July 21 - 27, 2024
-
August 11 - 16, 2024
-
February 15, 2024
In addition to its practical implications, recent work on “meaning representations” could shed light on some old philosophical questions.
-
April 16, 2024First model to work across a wide range of products uses a second U-Net encoder to capture fine-grained product details.
-
March 18, 2024Tokenizing time series data and treating it like a language enables a model whose zero-shot performance matches or exceeds that of purpose-built models.
-
February 20, 2024Generative AI supports the creation, at scale, of complex, realistic driving scenarios that can be directed to specific locations and environments.
-
January 17, 2024Representing facts using knowledge triplets rather than natural language enables finer-grained judgments.
-
ACL Findings 20242024We show that content on the web is often trans-lated into many languages, and the low quality of these multi-way translations indicates they were likely created using Machine Translation (MT). Multi-way parallel, machine generated content not only dominates the translations in lower resource languages; it also constitutes a large fraction of the total web content in those languages. We also find evidence
-
Transactions on Machine Learning Research2024Model miscalibration has been frequently identified in modern deep neural networks. Recent work aims to improve model calibration directly through a differentiable calibration proxy. However, the calibration produced is often biased due to the binning mechanism. In this work, we propose to learn better-calibrated models via meta-regularization, which has two components: (1) gamma network (γ-Net), a meta
-
Pixel-level mask annotation costs are a major bottleneck in training deep neural networks for instance segmentation. Recent promptable foundation models like the Segment Anything Model (SAM) and GroundedDINO (GDino) have shown impressive zero-shot performance in segmentation and object detection benchmarks. While these models are not capable of performing inference without prompts, they are ideal for omnisupervised
-
Interspeech 20242024End-to-end (E2E) automatic speech recognition (ASR) systems often exploited pre-trained hidden Markov model (HMM) systems for word timing estimation (WTE), due to their inability to predict word boundaries. However, training an HMM is difficult for low-resource languages due to the lack of phonetic transcriptions, leading to a high demand for HMM-free WTE methods, particularly for multilingual ASR systems
-
ACL Findings 20242024Speculative decoding has emerged as a powerful method to improve latency and through-put in hosting large language models. However, most existing implementations focus on generating a single sequence. Real-world generative AI applications often require multiple responses and how to perform speculative decoding in a batched setting while preserving its latency benefits poses non-trivial challenges. This
Resources
-
We look for talent from around the world for applied scientists, data scientists, economists, research scientists, scholars, academics, PhDs, and interns.
-
We hire world-class academics to work on large-scale technical challenges, while they continue to teach and conduct research at their universities. Learn more about each program and how to apply below.
-
Supporting research at academic institutions and non-profit organizations in areas that align with our mission to advance customer-obsessed science.