Spaces:
Running on Zero
Running on Zero
Add prompt repetition technique to Key Design Choices
Browse files
README.md
CHANGED
|
@@ -57,6 +57,7 @@ Four agents, each with a distinct adversarial role, orchestrated by [LangGraph](
|
|
| 57 |
- **Dual-source analysis** -- every agent considers both the medical image and clinical context (vitals, labs, risk factors), because many dangerous conditions have subtle imaging but obvious clinical red flags
|
| 58 |
- **MedSigLIP verification** -- zero-shot image classification grounds the bias analysis in visual evidence, not just language reasoning
|
| 59 |
- **MedASR voice input** -- [MedASR](https://huggingface.co/google/medasr) enables hands-free clinical context entry via speech-to-text, designed for busy clinical workflows where typing is impractical
|
|
|
|
| 60 |
- **Collegial tone** -- the Consultant writes as a consulting colleague (*"Have you considered..."*), not a critic
|
| 61 |
|
| 62 |
---
|
|
|
|
| 57 |
- **Dual-source analysis** -- every agent considers both the medical image and clinical context (vitals, labs, risk factors), because many dangerous conditions have subtle imaging but obvious clinical red flags
|
| 58 |
- **MedSigLIP verification** -- zero-shot image classification grounds the bias analysis in visual evidence, not just language reasoning
|
| 59 |
- **MedASR voice input** -- [MedASR](https://huggingface.co/google/medasr) enables hands-free clinical context entry via speech-to-text, designed for busy clinical workflows where typing is impractical
|
| 60 |
+
- **Prompt repetition** -- implements the [prompt repetition technique](https://arxiv.org/abs/2512.14982) from Google Research to improve output quality and consistency in non-reasoning LLMs
|
| 61 |
- **Collegial tone** -- the Consultant writes as a consulting colleague (*"Have you considered..."*), not a critic
|
| 62 |
|
| 63 |
---
|