Mark Chen

Overview

Mark Chen is a researcher at OpenAI. He joined the company in 2018 and, according to OpenAI Forum event biographies, has played a key role in its projects since then ^[2]. In March 2025, OpenAI's leadership update announced that he had moved into an expanded Chief Research Officer role ^[1].

Research contributions

Chen is credited with leading the team that created DALL·E 2 ^[2]. DALL·E 2 is a system that produces images from text descriptions. The technical paper describing its method, "Hierarchical Text-Conditional Image Generation with CLIP Latents", was submitted to arXiv on 13 April 2022, and Chen is one of its five authors ^[4]. The title names the approach: images are generated in stages, conditioned on text, using the latent representations of CLIP, a model that places images and text in a shared embedding space.

He is also credited with the development of Codex ^[2], OpenAI's work on models trained on source code. Chen is listed first among the authors of "Evaluating Large Language Models Trained on Code", submitted to arXiv on 7 July 2021 — a paper concerned with how large language models trained on code are evaluated ^[3]. Models of this kind are trained on large amounts of program text and then asked to produce code themselves; the paper's subject is how that ability is measured.

The two lines of work sit at different ends of the field: one generates images from language, the other generates and reasons about code. Both papers appeared under his name, one as first author and one as a co-author, which is how the public record of the work at publication time is preserved ^[3][4].

Remit and role

OpenAI Forum event biographies describe his remit as covering language models, reinforcement learning, multimodal models and AI alignment ^[2]. These are distinct kinds of work: language models are trained on text to predict and produce it; multimodal models handle more than one kind of data at once, such as text together with images; reinforcement learning trains a system with rewards rather than labelled examples; and alignment is the work of making a system's behaviour match what its users and builders intend.

The March 2025 leadership update gave the Chief Research Officer role an expanded scope. Alongside responsibility for scientific progress, the role was to tie research more closely to product work ^[1]. That pairing — research leadership that also carries responsibility for how research reaches products — is the arrangement the announcement described.

Place in the organisation

Chen's record at OpenAI covers work that became visible both as research papers and as systems people could use: DALL·E 2 for image generation and Codex for code ^[2][3][4]. His appointment to an expanded Chief Research Officer role in March 2025 placed him at the point where the company's research agenda and its product work meet ^[1], with the four areas named in his biography — language models, reinforcement learning, multimodal models and alignment — as the span of that responsibility ^[2].