---
task_type: credit_underwriting
license: mit
language:
- en
tags:
- lh-bench
- benchmark
- agents
- long-horizon
- rl-environment
- financial-services
- document-reasoning
- programmatic-verification
- tribal-knowledge
pretty_name: 'LH-Bench: Credit Underwriting (preview)'
size_categories:
- 1K<n<10K
---

# LH-Bench: Credit Underwriting Benchmark

> **Part of the LH-Bench suite** by Metaphi. The LH-Bench paper, [LH-Bench: Skill-Grounded Evaluation of Long-Horizon Agents on Subjective Enterprise Tasks](https://arxiv.org/abs/2603.22744), reports results on the Figma-to-code and programmatic-content environments; this credit-underwriting environment is released alongside them. If you use it, please cite the paper (citation at the end of this card).


    A 1387-task RL environment for financial document understanding. Agents are provided with business documents (bank statements), and asked 
    to assess the risk associated by using a context-rich classification taxonomy for each transaction . The agent's output is programmatically graded 
    against the ground truth labelled by domain epxerts with over 10+ years of experience in the risk assessment field. 

  ## Scale                                                                                                                   
                                                                                                                         
  - **1387 qualifying cases** (Businesses, loan-origination)                                                      
  - **5,607,517 total transactions**                
  - **1-73 PDFs per case** (median: 4, mean: 6.4)                                                                            
    

  ## Task Definition                                                                                                         
                                                                                                                         
  | | |                                                                                                                      
  |---|---|                                                                                                                  
  | **Input** | 1-73 bank statement PDFs per case, no PII information |                                                                          
  | **Output** | Extract Structured JSON: transactions (date, amount, balance, narration) and categorize transaction type, counterparty |                                                                                                             
  | **Reward** | Extraction F1 (hash-based), categorization accuracy |                                                       
  | **Grading** | Fully programmatic, instant (~seconds per case) |

  ## Difficulty Distribution

  ### By PDF Count                                                                                                           
                                                                                                                             
  | PDFs | Cases | % of Total |                                                                                              
  |------|-------|-----------|
  | 1 | 269 | 19.4% |
  | 2-3 | 406 | 29.3% |
  | 4-6 | 317 | 22.9% |
  | 7-10 | 154 | 11.1% |
  | 10+ | 241 | 17.4% |

  ### By Transaction count

  | Transactions | Cases | % of Total |                                                                                      
  |-------------|-------|-----------|
  | <100 | 11 | 0.8% |                                                                                                       
  | 100-500 | 153 | 11.0% |                                                                       
  | 500-2,000 | 535 | 38.6% |                                                                                                
  | 2,000+ | 688 | 49.6% |  


 ### Financial Data Analyst (Extraction)                                                                                    
                                                                                                                             
  Extracts structured transaction data from PDF bank statements.

  | Model | Harness | Pass@1 | 95% CI | Tasks |
  | --- | --- | --- | --- | --- |
  | GPT-5.4 | Codex | 33.3% | 8.3 - 58.3% | 12 |
  | Claude Opus 4.6 | Claude Code | 31.3% | 12.5 - 56.3% | 16 |
  | Gemini 3.1 Pro | Gemini CLI | 12.5% | 0.0 - 29.2% | 12 |

  ### Credit Analyst (Categorization)

  Categorizes bank transactions by type, counterparty, and group.

  | Model | Harness | Pass@1 | 95% CI | Tasks |
  | --- | --- | --- | --- | --- |
  | Claude Opus 4.6 | Claude Code | 31.3% | 12.5 - 56.3% | 16 |
  | GPT-5.4 | Codex | 16.7% | 0.0 - 41.7% | 12 |
  | Gemini 3.1 Pro | Gemini CLI | 12.5% | 0.0 - 33.3% | 12 |
  

  ## For RL Training                                                                                                         
                                                                                                                             
  - **Reward signal**: Extraction F1 as scalar reward (0.0 to 1.0), fully programmatic                                       
  - **Rollout**: Agent receives PDFs → produces structured JSON → verifier computes F1 in seconds                            
  - **Parallelizable**: Each case is independent — runs in an isolated container                                             
  - **Difficulty calibration**: Easy (84% F1) to hard (43% F1) — significant headroom                                        
  - **No human evaluation needed**: reward signal is instant and deterministic
  - **Decomposed rewards available**: extraction F1 + categorization accuracy + counterparty accuracy as separate signals

  ## Dataset Schema

    ```json
    {
      "id": "case-000",
      "case_id": "uuid-based-case-identifier",
      "pdf_gcs_uris": ["gs://.../_scrubbed.pdf"],
      "pdf_filenames": ["fileid_scrubbed.pdf"],
      "extraction_ground_truth": "{...}",
      "categorization_ground_truth": "{...}",
      "description": "3 PDFs - 792 txns",
      "case_type": "Business Case",
      "gt_status": "clean"
    }
    ```

## Leaderboard

Results, task sets and every LH-Bench dataset: [metaphi.ai/benchmarks/lh-bench](https://metaphi.ai/benchmarks/lh-bench). Benchmark index on GitHub: [metaphi-labs/lh-bench](https://github.com/metaphi-labs/lh-bench). All datasets in one place: [LH-Bench collection](https://huggingface.co/collections/metaphilabs/lh-bench-6acd48af30ec71aff2a0ffcd).

## Paper and citation

Paper: [LH-Bench: Skill-Grounded Evaluation of Long-Horizon Agents on Subjective Enterprise Tasks](https://arxiv.org/abs/2603.22744) (arXiv:2603.22744).

```bibtex
@article{chandwani2026lhbench,
  title   = {LH-Bench: Skill-Grounded Evaluation of Long-Horizon Agents on Subjective Enterprise Tasks},
  author  = {Chandwani, Abhishek and Gupta, Ishan},
  journal = {arXiv preprint arXiv:2603.22744},
  year    = {2026}
}
```

## Other LH-Bench datasets

- [Figma-to-Code tasks](https://huggingface.co/datasets/metaphilabs/frontend-figma-to-code)
- [Figma-to-Code expert preferences](https://huggingface.co/datasets/metaphilabs/figma2code-expert-preferences)
- [Programmatic Content tasks](https://huggingface.co/datasets/metaphilabs/remotion-video-gen)
- [Programmatic Content expert preferences](https://huggingface.co/datasets/metaphilabs/remotion-video-preferences)
- [Credit Underwriting (commercial, gated)](https://huggingface.co/datasets/metaphilabs/credit-underwriting-commercial)