Model evidence / current era
Recorded agent setup
Tencent HY4 (preview)
Effort: think · current era
Harness: Claude Code
Task selection is incompletely recorded; catalog-wide generalization is unsupported.
20.0%macro pass estimate
3eligible benchmark task versions
17valid runs on those tasks
8HARNESS_ERROR · excluded in scope
Macro: 20.0% [0.0%, 60.0%] · Micro: 3/17 · 17.6% [6.2%, 41.0%]. Latest nondefective task versions, each with ≥3 valid runs; a macro needs ≥3 tasks. The model’s measured subset differs from other rows.
Domain results
| Domain | Tasks / runs | Macro [95% bootstrap] | Micro [95% Wilson] | Calibration |
|---|---|---|---|---|
| Cloud operations | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Data processing / data science | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| DeepSWE (long-horizon SWE) | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Electrical engineering (RTL) | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Games | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Machine learning | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Mechanical engineering | 1 / 4 | Insufficient task coverage | 0/4 · preliminary | Limited coverage |
| Media | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Operations | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Other / uncategorised | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| SWE-Marathon | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Science / STEM | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Scientific computing | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Security / cyber | 1 / 5 | Insufficient task coverage | 3/5 | Limited coverage |
| Snorkel-reviewed tasks (category unknown) | 0 / 0 | Insufficient task coverage | 0/0 | Limited coverage |
| Software engineering | 1 / 8 | Insufficient task coverage | 0/8 | Limited coverage |
Resource coverage
Valid runs on nondefective benchmark versions. Missing values are excluded, never counted as zero. Medians are descriptive and can reflect different task difficulty and harness behavior.
| Recorded resource | Median | Observed / eligible runs |
|---|---|---|
| Recorded tokens | 102,237 | 23 / 23 |
| Cost (USD) | 7.4 | 23 / 23 |
| Agent turns | 75 | 23 / 23 |
| Duration (seconds) | 5,143.72 | 8 / 23 |
Task-level evidence
Includes excluded and superseded versions for inspection; only the chosen nondefective versions enter aggregate results. Each task links to the exact runs and their diagnostic provenance.
| Task / version | Domain | Verifier outcome [95% Wilson] | Tier / aggregate scope |
|---|---|---|---|
| sqlite-distinct-agg-arityVersion 2 | DeepSWE (long-horizon SWE) | 4/4 · 100.0% [51.0%, 100.0%] · preliminary | contested/defectiveExcluded / other version |
| faiss-scan-exactly-max-codes-elements-26Version 3 | DeepSWE (long-horizon SWE) | 0/1 | contested/defectiveExcluded / other version |
| qdrant-fix-wrong-multivectors-count-in-sVersion 2 | DeepSWE (long-horizon SWE) | 0/0 | contested/defectiveExcluded / other version |
| qdrant-fix-wrong-multivectors-count-in-sVersion 4 | DeepSWE (long-horizon SWE) | 3/8 · 37.5% [13.7%, 69.4%] | contested/defectiveExcluded / other version |
| rtl-axi-event-snapshot-bankVersion 1 | Electrical engineering (RTL) | 2/2 | uncalibratedBenchmark version |
| landmark-competing-risk-scorecardVersion 1 | Machine learning | 0/1 | uncalibratedBenchmark version |
| cam-modal-toolpath-envelopeVersion 1 | Mechanical engineering | 0/1 | uncalibratedBenchmark version |
| dxf-nested-assembly-scheduleVersion 1 | Mechanical engineering | 0/4 · 0.0% [0.0%, 49.0%] · preliminary | uncalibratedBenchmark version |
| family-benefit-accumulator-reprocessingVersion 1 | Operations | 0/1 | contested/defectiveExcluded / other version |
| occurrence-reinsurance-recovery-bordereauVersion 1 | Operations | 0/0 | uncalibratedBenchmark version |
| webauthn-assertion-transaction-gateVersion 1 | Security / cyber | 1/1 | uncalibratedBenchmark version |
| tuf-offline-update-chainVersion 1 | Security / cyber | 0/1 | uncalibratedBenchmark version |
| oci-layer-admission-auditVersion 1 | Security / cyber | 3/5 · 60.0% [23.1%, 88.2%] | mediumBenchmark version |
| etcd-fix-race-berween-read-index-and-leaVersion 2 | Software engineering | 0/8 · 0.0% [0.0%, 32.4%] | hardBenchmark version |
| rocksdb-fix-compaction-abort-reschedulinVersion 2 | Software engineering | 4/4 · 100.0% [51.0%, 100.0%] · preliminary | contested/defectiveExcluded / other version |
| landmark-competing-risk-scorecardVersion 4 | Machine learning | 0/8 · 0.0% [0.0%, 32.4%] | contested/defectiveExcluded / other version |
| occurrence-reinsurance-recovery-bordereauVersion 4 | Operations | 0/8 · 0.0% [0.0%, 32.4%] | contested/defectiveExcluded / other version |
| oci-layer-admission-auditVersion 1 | Security / cyber | 0/7 · 0.0% [0.0%, 35.4%] | contested/defectiveExcluded / other version |
| openssh-certificate-authorizationVersion 1 | Security / cyber | 3/8 · 37.5% [13.7%, 69.4%] | contested/defectiveExcluded / other version |