Doubao Seed 2.0 Lite: benchmark results
Provider: ByteDance. Access: Open.
Unified ELO 1778 ± 22, rank #81 of 1629 rated models, from 18 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ViSTR-Bench | 58.8 | Accuracy (%; averaged over all 1,340 two-option questions in | 94.7 |
| FoodMonitor - Environment Violations | 42.1 | Environment-violation F1 (0-100; the paper's 0-1 score times | 90 |
| MMGist | 55 | Macro ↑ (self-reported) | 84.6 |
| FoodMonitor | 29.8 | C score (0-100; the paper's 0-1 score times 100): the mean o | 80 |
| PortBench-QA - Rebalancing | 81 | Item score (%; rebalance-or-hold decision with the correctiv | 77.8 |
| PortBench-QA | 80.4 | Mean item score (%; mean of seven QA templates, 50 test ques | 66.7 |
| LisanBench | 0.06 | Mean Path Length / Current Maximum | 66 |
| ToolPrivacyBench - Public-derived - Task Success | 87.77 | TaskSuccess (%) | 62.5 |
| ToolPrivacyBench - Synthetic-private - Forbidden Disclosure | 21.67 | Forbidden Over-disclosure Rate (%) | 62.5 |
| ToolPrivacyBench - Synthetic-private - Severity-weighted Leakage | 23.3 | Severity-weighted Leakage Rate (%) | 62.5 |
| FoodMonitor - Person Violations | 17.5 | Person-violation F1 (0-100; the paper's 0-1 score times 100) | 60 |
| ToolPrivacyBench - Public-derived - Severity-weighted Leakage | 13.93 | Severity-weighted Leakage Rate (%) | 50 |
Interactive version: theaggregate.ai/model?slug=doubao-seed-2-0-lite · How It Works · Data refreshed daily, snapshot 2026-10-07.