Dianjin-R1-7B: benchmark results
Provider: Other.
Aggregate Rating 1483 ± 33, rank #2629 of 10916 rated models, from 39 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FinRiskAtlas - Evidence-Grounded Processing - Institution Matching | 90 | Accuracy (%) | 87.5 |
| QuantEval - Quantitative Reasoning (No CoT) | 38 | Accuracy (%; direct answer, no chain-of-thought) | 83.3 |
| QuantEval - Quantitative Reasoning (CoT) | 42.8 | Accuracy (%; chain-of-thought prompting) | 75 |
| FinRiskAtlas - FinRisk-Ask - Evidence-Request Alignment | 57.72 | ERA (%; DeepSeek-V4-Flash request-alignment judge) | 54.8 |
| FinED-Bench - Financial Domain Knowledge Errors | 17.19 | F1 (%; financial domain knowledge errors: terminology misuse | 40 |
| FinED-Bench - Financial Reasoning Errors | 8.66 | F1 (%; financial reasoning errors: conflicting expression, t | 40 |
| FinReasoning (Deep Insight) | 70.8 | Rubric score (0-100): a judge score fused from DeepSeek-V3 a | 33.3 |
| FinReasoning (Deep Insight) - Architectural Richness and Strictness | 77.1 | Rubric score (0-100): a judge score fused from DeepSeek-V3 a | 33.3 |
| FinReasoning (Deep Insight) - Completeness and Comparative Analysis | 73.3 | Rubric score (0-100): a judge score fused from DeepSeek-V3 a | 33.3 |
| FinReasoning (Deep Insight) - Justification and Causal Depth | 69.1 | Rubric score (0-100): a judge score fused from DeepSeek-V3 a | 33.3 |
| FinED-Bench - General Knowledge Errors | 18.72 | F1 (%; general knowledge errors: illegal time, redundant sta | 30 |
| FinReasoning (Deep Insight) - Factuality and Contextualization | 63.6 | Rubric score (0-100): a judge score fused from DeepSeek-V3 a | 27.8 |
Interactive version: theaggregate.ai/model?slug=dianjin-r1-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.