Michael Wu

Aaron Hurst, Adam Lerer, Adam P. Goucher et al. · openai

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

Josh Achiam, Steven Adler, Sandhini Agarwal et al. · berkeley, deepmind

We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4.

Chao Peter Yang, Zixiao Tan, Kaisen Yao et al.

Multi-agent systems powered by large language models (LLMs) are increasingly used for financial analysis and decision support. However, existing coordination schemes, especially those emphasizing consensus or debate, are vulnerable to sycophancy: agents conform to peer reasoning instead of evidence, leading to premature agreement and degraded outcomes. We introduce FinCom (Financial Committee), a governed multi-agent framework and interactive system that operationalizes the Disagree-or-Commit (DoC) protocol to embed structured dissent into financial AI committees. A central Supervisor orchestrates three ReAct-enabled specialist agents: Research, Quantitative, and Risk. Each agent is equipped with role-specific tools for retrieval, computation, and stress testing. During deliberation, agents must either explicitly critique or commit to their peers' reasoning before converging on a unified recommendation. This demonstration showcases how FinCom supports committee-style financial analysis through coordinated multi-agent interaction, including structured report generation and interactive decision support. Evaluated across the most recent financial agent benchmark, in addition to 90 internal handcrafted financial tasks using an LLM-as-a-Judge protocol, DoC improves reasoning accuracy and risk awareness significantly over a consensus-seeking baseline on both an in-house and external evaluation set. By reframing disagreement as a governance primitive rather than noise, FinCom offers a lightweight, prompt-only recipe for improving accountability, transparency, and epistemic robustness in agentic financial systems.

Sambhav R. Jain, Albert Gural, Michael Wu et al.

We propose a method of training quantization thresholds (TQT) for uniform symmetric quantizers using standard backpropagation and gradient descent. Contrary to prior work, we show that a careful analysis of the straight-through estimator for threshold gradients allows for a natural range-precision trade-off leading to better optima. Our quantizers are constrained to use power-of-2 scale-factors and per-tensor scaling of weights and activations to make it amenable for hardware implementations. We present analytical support for the general robustness of our methods and empirically validate them on various CNNs for ImageNet classification. We are able to achieve near-floating-point accuracy on traditionally difficult networks such as MobileNets with less than 5 epochs of quantized (8-bit) retraining. Finally, we present Graffitist, a framework that enables automatic quantization of TensorFlow graphs for TQT (available at https://github.com/Xilinx/graffitist ).

Belal Shaheen, Minh-Hieu Nguyen, Bach-Thuan Bui et al.

Aerial remote sensing enables efficient large-area surveying, but accurate direct object-level measurement remains difficult in complex natural scenes. Recent advancements in 3D vision, particularly learned radiance-field representations such as NeRF and 3D Gaussian Splatting, have begun to raise the ceiling on reconstruction fidelity and densifiable geometry from posed imagery. Nevertheless, direct aerial measurement of important natural attributes such as tree diameter at breast height (DBH) remains challenging. Trunks in aerial forest scans are distant and sparsely observed in image views: at typical operating altitudes, stems may span only a few pixels. With these constraints, conventional reconstruction methods leave breast-height trunk geometry weakly constrained. We present TreeDGS, an aerial image reconstruction method that leverages 3D Gaussian Splatting as a continuous, densifiable scene representation for trunk measurement. After SfM-MVS initialization and Gaussian optimization, we extract a dense point set from the Gaussian field using RaDe-GS's depth-aware cumulative-opacity integration and associate each sample with a multi-view opacity reliability score. We then estimate DBH from trunk-isolated points using opacity-weighted solid-circle fitting. Evaluated on 10 plots with field-measured DBH, TreeDGS reaches 4.79,cm RMSE (about 2.6 pixels at this GSD) and outperforms a state-of-the-art LiDAR baseline (7.91,cm RMSE), demonstrating that densified splat-based geometry can enable accurate, low-cost aerial DBH measurement.

Mohammad Keshavarzi, Michael Wu, Michael N. Chin et al.

Virtual and augmented reality communication platforms are seen as promising modalities for next-generation remote face-to-face interactions. Our study attempts to explore non-verbal communication features in relation to their conversation context for virtual and augmented reality mediated communication settings. We perform a series of user experiments, triggering nine conversation tasks in 4 settings, each containing corresponding non-verbal communication features. Our results indicate that conversation types which involve less emotional engagement are more likely to be acceptable in virtual reality and augmented reality settings with low-fidelity avatar representation, compared to scenarios that involve high emotional engagement or intellectually difficult discussions. We further systematically analyze and rank the impact of low-fidelity representation of micro-expressions, body scale, head pose, and hand gesture in affecting the user experience in one-on-one conversations, and validate that preserving micro-expression cues plays the most effective role in improving bi-directional conversations in future virtual and augmented reality settings.

Sean O. Settle, Manasa Bollavaram, Paolo D'Alberto et al.

Deep learning as a means to inferencing has proliferated thanks to its versatility and ability to approach or exceed human-level accuracy. These computational models have seemingly insatiable appetites for computational resources not only while training, but also when deployed at scales ranging from data centers all the way down to embedded devices. As such, increasing consideration is being made to maximize the computational efficiency given limited hardware and energy resources and, as a result, inferencing with reduced precision has emerged as a viable alternative to the IEEE 754 Standard for Floating-Point Arithmetic. We propose a quantization scheme that allows inferencing to be carried out using arithmetic that is fundamentally more efficient when compared to even half-precision floating-point. Our quantization procedure is significant in that we determine our quantization scheme parameters by calibrating against its reference floating-point model using a single inference batch rather than (re)training and achieve end-to-end post quantization accuracies comparable to the reference model.