EBITDA Add-Back Identification
Given a synthetic general ledger and normalization policy, identify candidate nonrecurring expenses, calculate proposed add-backs, and provide the policy basis for each.
View token rates · observed 9/11/2026
Compare 2 router rates · observed 9/11/2026
Compare 2 router rates · observed 9/11/2026
All 12 model results and methodology
Generated 100/100 examples
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 26% | 100/100 | $50.25 |
| Qwen3 235B A22B | mid | 52% | 100/100 | $382 |
| DeepSeek V3 | mid | 69% | 100/100 | $183 |
| Mistral Large 2407 | mid | 50% | 100/100 | $322 |
| Arcee Trinity Large Thinking | mid | 36% | 100/100 | $171 |
| GPT-5.4 | frontier | 70% | 100/100 | $3008 |
| Claude Opus 4.7 | frontier | 55% | 100/100 | $5684 |
| Gemini 3.1 Pro Preview | frontier | 0% | 99/100 | $2598 |
| Gemma 4 E4B IT | small | 42% | 100/100 | $18.22 |
| Granite 4.1 8B | small | 33% | 100/100 | $18.43 |
| Ministral 8B Instruct 2410 | small | 17% | 100/100 | $34.47 |
| Qwen3 4B Instruct 2507 | small | 26% | 100/100 | $133 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-10T09:00:02.785Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock