PeopleofSuperintelligence

Researcher

Bin Wang

Bin Wang contributes infrastructure and model-systems work to StepFun's AI projects. He works on real-time speech interaction and audio-language models, and coauthors work on cost-aware decoding and continuous-token image generation. His work also includes sparse language models and coordinated reasoning. As of October 2026, he contributes to PaCoRe's infrastructure operations.

Works
18
Citations
76
h-index
5

Publication numbers from OpenAlex, as of 19 October 2026.

Timeline

  1. February 2025

    Real-time speech interaction

    In February 2025, he contributes to Step-Audio, a production-ready open-source system for real-time speech interaction.

  2. June 2025

    Infrastructure for audio answers

    In June 2025, he contributes infrastructure work to Step-Audio-AQAA, a fully end-to-end model for audio query-to-answer tasks with a 130-billion-parameter backbone.

  3. July 2025

    Step-Audio 2

    In July 2025, he coauthors Step-Audio 2, an end-to-end multimodal language model for audio understanding and speech conversation.

  4. July 2025

    Cost-aware Step-3 decoding

    In July 2025, he coauthors Step-3's model-system report for cost-effective decoding, which presents a 321-billion-parameter vision-language model.

  5. 2025

    Continuous-token image generation

    In 2025, he coauthors NextStep-1, a 14-billion-parameter autoregressive image-generation model that uses continuous image tokens.

  6. February 2026

    A sparse agentic model

    In February 2026, he coauthors Step 3.5 Flash, a sparse mixture-of-experts model that activates 11 billion of its 196 billion parameters per token.

  7. As of October 2026

    PaCoRe infrastructure operations

    As of October 2026, he is an infrastructure-operations contributor to PaCoRe, a parallel coordinated reasoning project supported by StepFun and Tsinghua University.

Most-cited works

  1. An Information Hiding Algorithm for HEVC Based on Differences of Intra Prediction Modes2016 · Lecture notes in computer science · 18 citations
  2. Real-time sound source localization using hybrid framework2013 · Applied Acoustics · 18 citations
  3. Enhanced adaptive subtraction method for simultaneous source separation2014 · Second International Meeting for Applied Geoscience & Energy · 9 citations
  4. Generalized Channel Coding and Decoding With Natural Redundancy in Protocols2023 · IEEE Transactions on Communications · 6 citations
  5. Multihydrophone Fusion Network for Modulation Recognition2022 · Sensors · 6 citations

From OpenAlex, as of 19 October 2026.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI from 1961 to 2025, with a year-by-year chart.
  2. The United States employs an estimated 53% of people in our data with a known employment countryOur estimates of where AI careers are concentrated now and how countries’ shares of dated positions changed over time.
  3. 46% of living people in our data with known birth and employment countries work outside their birth countryOur estimate of cross-border careers in AI, from country of birth and the country of each current organization.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionOur estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.

Sources and credits

Photo: We have no published photo of Bin Wang yet. Send us a free-licensed photo.

  1. [2502.11946] Step-Audio: Unified Understanding and Generation in Intelligent Speech Interaction arxiv.org. Checked 18 October 2026.
  2. [2506.08967] Step-Audio-AQAA: a Fully End-to-End Expressive Large Audio Language Model arxiv.org. Checked 18 October 2026.
  3. [2507.16632] Step-Audio 2 Technical Report arxiv.org. Checked 18 October 2026.
  4. [2507.19427] Step-3 is Large yet Affordable: Model-system Co-design for Cost-effective Decoding arxiv.org. Checked 18 October 2026.
  5. GitHub - stepfun-ai/NextStep-1: [🚀 ICLR 2026 Oral] NextStep-1: SOTA Autogressive Image Generation with Continuous Tokens. A research project developed by the StepFun’s Multimodal Intelligence team. · GitHub github.com. Checked 18 October 2026.
  6. [2602.10604] Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters arxiv.org. Checked 18 October 2026.
  7. stepfun-ai/Step-3.5-Flash · Hugging Face huggingface.co. Checked 18 October 2026.
  8. PaCoRe/README.md at main · stepfun-ai/PaCoRe · GitHub github.com. Checked 18 October 2026.