Koutian Wu
23 yo UT-Austin PhD-dropout researcher
@ktwu01 1,777 X 追蹤者 ut01 ktwu01ktwu01.github.io 首個產品發布於 18 天前
- 公開營收
- 未公開
- 名下產品暫無公開營收
名下產品
| 產品 | 每月造訪次數 · 10月 | 操作 |
|---|---|---|
| ut01開源面向德克薩斯大學奧斯汀分校學生的導航資源頁,匯總常用校園服務與資源連結。 | <5K | |
| Benchmark Radar開源面向 AI 研究者和評測構建者的資料庫與搜尋引擎,用於查找基準測試及其論文、程式碼、資料集和模型分數。 | <5K |
時間線
11 月 18 日 · 5 天前
Benchmark Radar releases v0.15.0產品動態
Benchmark Radar v0.15.0 adds faster search, more benchmarks and reported scores, improved share cards, and the Cohere Labs talk deck.
11 月 17 日 · 6 天前
Benchmark Radar covers the AgentShield multi-turn attack benchmark引薦與反向連結
Benchmark Radar reports on the AgentShield Multi-Turn Agent Attack Benchmark, which tests whether security frameworks prevent unauthorized tool execution across multi-turn adversarial interactions. The report also mentions separate benchmarks for pagination, retries, and cancellation.
Benchmark Radar covers Tiny Aya Compression Evaluation引薦與反向連結
Benchmark Radar shares Tiny Aya Compression Evaluation by Mudit666. The dataset contains per-question and per-prompt scores for two studies of CohereLabs/tiny-aya-global under 4-bit compression, examining unlearning and refusal behaviour.
Benchmark Radar covers HealthRouter Bench引薦與反向連結
Benchmark Radar covers HealthRouter Bench, an evaluation dataset for per-sample routing over frozen medical image encoders. The description states that ten encoders are evaluated on four pooled medical tasks and that dataset access is gated and manually approved.
Benchmark Radar covers LLM Benchmarks Losing Their Judge引薦與反向連結
Benchmark Radar shares Ogous's study of Arena-Hard v2, SimpleQA and HELM WildBench. The study examines how verdicts change when the LLM judge is replaced while grading the same stored answers, with call records, anchorset files and recomputation scripts provided in the dataset release.
Benchmark Radar covers SiteShift prostate tumour segmentation benchmark引薦與反向連結
Benchmark Radar shares SiteShift, a cross-site benchmark for prostate tumour segmentation. The description says it tests model performance at hospitals not represented in training data and includes prostate H&E segmentation data from Central Asia; access requests are not processed before publication.
Benchmark Radar covers the MLIRBench benchmark引薦與反向連結
Benchmark Radar lists MLIRBench as a new benchmark for reasoning about Multi-Level Intermediate Representation, a compiler format used to express and transform programs.
Benchmark Radar covers the HydroTuring benchmark引薦與反向連結
Benchmark Radar lists HydroTuring as a new black-box benchmark with 37 probes for mass, energy, and momentum conservation in hydrologic models.
Benchmark Radar covers a study of financial research agent reliability引薦與反向連結
Benchmark Radar describes a study that reruns one shipped financial-research agent seven times on 18 questions. Identical runs satisfy between 105 and 123 of 149 rubric criteria, while the reported bill varies by as much as 17%; the study also assigns sampled failures to instruction, tool, model, or data owners.
Benchmark Radar covers a study of verification transport in agent evaluation引薦與反向連結
Benchmark Radar shares a study introducing matched native, abstracted, and restored views of the same agent execution. The test examines whether a language-model verifier maintains its judgment when tool composition, delegation, or history compaction changes the visible evidence.
問題與證據
- Koutian Wu 開發過哪些產品?
- EarlyLore 收錄了 Koutian Wu 的 2 款獨立軟體產品(其中 2 款當前線上),代表產品包括 ut01, Benchmark Radar。
- Koutian Wu 旗下產品的網站造訪次數有多少?
- 按各獨立網域名稱最近一個可用統計月份合計,Koutian Wu 旗下產品可追蹤的網頁每月造訪次數為 5.1K;共享託管網域名稱與商店擴展頁面不計入該合計。
相關獨立開發者研究指南