Skip to content

News ​

Each result gets its own post, with the number and its caveats in the same place, dated, with the people who produced it named at the top and a link to whatever it can be checked against. The longer arguments are on the blog. Subscribe by RSS.

October 2026

  1. HOA#1 on Lean-Eval v1 first solves

    Lean-Eval froze its first 128-problem release and moved everything older to an archive. On v1 our GPT-5.6 flow has the most first solves, 30, and is one behind the leader on total; on the archive it has 170 of 171.

    Zhengyang Zhang, Hongzhou Lin, Jui-Hui Chung, Zijian Zhang
  2. HMA78.2% MLE-bench medal rate in 6 hours

    Two coding agents taking turns over one workspace, each starting fresh, beat both of them working alone on 75 MLE-bench tasks: 78.2% any-medal against 72.4% and 68.0%, in the same six hours. Self-reported, and not on the official leaderboard.

    Changye Li, Zijian Zhang
  3. KDA6 KDA PRs merged into SGLang

    Since August, SGLang has merged six pull requests carrying kernels our agents wrote, the best adding 8.7% end-to-end throughput. The MLSys contest kernels are public with a benchmark anyone can rerun. One merge elsewhere, Flash-KMeans, was reverted the next day.

    Xiaoyu Zhang, Dongyun Zou, Ligeng Zhu, Zijian Zhang

September 2026

  1. HMABiohub final: #188 of 3,947, top 5%

    In August an agent workflow sat 8th on the Biohub cell-tracking public leaderboard. The final private rank is 188th of 3,947, the top 4.76%: lower, but an official finish. Plus the Kaggle tally when only final ranks count.

    Changye Li, Zijian Zhang
  2. HOA30/30 IPhO, 6/6 IOI, 100/100 IBO

    IPhO 2026 theory 30 of 30, IOI 2026 six of six, IBO 2024 theory 100 of 100, IChO 2026 formalized 68 of 68, and quantum information theory 40 of 40. What each score is, who graded it, and what it does not show.

    Jing Xiong, Zhengyang Zhang, Ligeng Zhu, Zijian Zhang

August 2026

  1. HOA672/672 on PutnamBench, joint #1

    GPT-5.6 in a worker–reviewer loop proved all 672 PutnamBench statements in Lean, at $44.50 a problem. The top of the official leaderboard is now a four-way tie, and what separates the four is cost.

    Ligeng Zhu, Zhengyang Zhang, Zijian Zhang
  2. HOA#1 on Lean-Eval: 172 proofs

    Three weeks after second place, a fully agentic GPT-5.6 run is first on the Lean-Eval leaderboard, with 172 research-level problems solved. Every accepted proof is sorry-free and independently re-verified.

    Zhengyang Zhang, Hongzhou Lin, Zijian Zhang
  3. HMATwo Kaggle finishes in the top 3%

    Ten agent workflows on Kaggle, with every result sorted by what kind of rank it is. Two closed with official final ranks in the top 3%. Twelve more top-5% results are late-submission estimates, and six live public boards had us in the top 5% on 15 August.

    Changye Li, Menghan Li, Yitong Liu, Zijian Zhang
  4. KDA6.5× faster MSA prefill indexer

    On B300, KDA found the hardware the MSA prefill and decode indexers were leaving idle: a 6.5× geometric mean on prefill, up to 14× on chunk tails, up to 3.3× on decode, and bitwise-identical output throughout. Humanfia-reported.

    Jiaming Tang, Zijian Zhang
  5. Humanize0% + 0.5% → 3.5% on ProgramBench

    Alone, Opus-4.8 fully resolves 0% of ProgramBench and GPT-5.5 resolves 0.5%. As a builder and a reviewer in one loop they resolve 3.5%, close to the best official entry's 4.5%, and the curve was still rising when the four-hour budget ran out. A Humanfia-reported checkpoint.

    Zheng Du
  6. KDAPast human SOTA on all 3 MLSys tracks

    Rerun on the MLSys 2026 FlashInfer contest, KDA's second-generation kernels beat the best human entry on every track: 1.69× on GDN prefill, 1.41× on DSA and 1.17× on MoE, in a public benchmark that pins the human winners' code.

    Dongyun Zou, Zijian Zhang

July 2026

  1. Humanize17,520 QCode definitions across 21 lattices

    QCode Discovery has searched 17,520 quantum-code definitions over 21 lattices and kept 5,178 that could win. Now nothing counts as a win without an exact construction, an independent check and a fresh replay of the IBM baselines. Humanfia-reported.

    Jing Xiong
  2. HOA23/23 at IPhO 2026, in Lean

    All 23 IPhO 2026 theory subproblems formalized and proved in Lean 4, sorry-free. Alongside: 36 of 36 on a blind quantum-algorithms benchmark, and 92.5% end to end on quantum information theory.

    Jing Xiong, Zhengyang Zhang, Zijian Zhang
  3. HOA#2 on Lean-Eval, behind Seed Prover

    A general model in a general agent loop is second on Lean-Eval, behind ByteDance's Seed Prover, a system built for formal mathematics. By our count it is 149 problems to 155.

    Zhengyang Zhang, Zijian Zhang
  4. HOA6/6 at IMO 2026, Lean-checked

    A fully agentic run solved all six problems of the 2026 International Mathematical Olympiad, every proof machine-checked in Lean 4. Two workers did it independently, the faster in 3.0× less API time than AxiomProver.

    Zhengyang Zhang, Zijian Zhang
  5. KDA#1 on SOL-ExecBench L1

    KDA-generated kernels put the Databricks entry first on SOL-ExecBench's L1 single-operation track, at a SOL score of 0.7639. It is still first on that board, by 0.00006.

    Lesheng Jin, Yuchen Jin, Ligeng Zhu, Dongyun Zou, Zijian Zhang

June 2026

  1. HOA670/672 on PutnamBench

    One Ralph loop, with Codex as both worker and reviewer, closed 670 of PutnamBench's 672 formal statements, every problem of Putnam 2025 among them. The best on the leaderboard that day was 668.

    Zhengyang Zhang, Zijian Zhang
  2. KDA2 KDA PRs merged in AMD's FlyDSL

    On AMD MI350X, with thin documentation and few published kernels to copy, KDA landed two pull requests in FlyDSL: a batch-aware routing fix worth up to 8% on GQA attention, and a correct fp8 attention forward path.

    Jin Pan, Zijian Zhang
  3. KDA71 first places on SOL-ExecBench

    KDA-generated kernels on the Databricks entry hold 71 of SOL-ExecBench's 235 per-kernel first places, more than any other entrant. It started in June with one 8×B200 node running unattended for a week.

    Lesheng Jin, Yuchen Jin, Zijian Zhang
  4. KDA40+ agent-tuned operators in SGLang

    More than forty agent-optimised operators are upstream in SGLang. Six of them, with their own pull requests' numbers: +71% throughput, −41% time to first token, 1.52× KDA prefill, up to 2.78× denoising and a 1.41× VAE decode.

    Xiaoyu Zhang, Zijian Zhang

May 2026

  1. KDATop 3 on every MLSys 2026 track

    KDA placed #1 on MoE, #2 on DSA and #3 on GDN in the Full-Agent tracks of the MLSys 2026 FlashInfer contest. Its best kernel ran 19.1× faster than the FlashInfer baseline, and on two of the five it was still slower.

    Dongyun Zou, Zhekai Zhang, Shang Yang, Hao Kang, Zijian Zhang

March 2026

  1. RLCR FlowA 567-file CMake port of gem5

    One engineer and the RLCR Flow loop took gem5 off SCons: CMake + Ninja as the build, Bazel alongside it, across 567 files, 133 commits and 42,967 changed lines. It is upstream as a request for comments.

    Sihao Liu