license: cc-by-nc-sa-4.0
language:
- en
- hi
base_model: Qwen/Qwen3-0.6B
pipeline_tag: text-generation
library_name: transformers
tags:
- lora
- peft
- qwen
- agriculture
KLM-1
KLM-1 is a bilingual agricultural language model supporting Hindi and English.
Base Model
Qwen3-0.6B
Evaluation
Benchmark: bharatgenai/BhashaBench-Krishi
Benchmark Results
| Model | BBK_English | BBK_Hindi |
|---|---|---|
| Llama-3.2-1B | 29.71 | 25.21 |
| Llama-3.2-1B-Instruct | 29.16 | 26.33 |
| Llama-3.2-3B | 32.68 | 28.69 |
| Granite-3.1-3b-a800m-base | 33.36 | 26.70 |
| Sarvam-2B-v0.5 | 28.14 | 25.57 |
| Sarvam-1-2B | 30.82 | 27.57 |
| BharatGen-AgriParam (2.9B) | 33.10 | 27.97 |
| SoketAI-Sarthi-Agri-v1 (27B) | 29.19 | 28.15 |
| KLM-1 (0.6B) | 47.11 | 38.52 |
| KLM-2 (0.6B) | 60.59 | 48.89 |
Intended Use
KLM-1 is intended for research, experimentation, and academic evaluation of bilingual agricultural language models. The model may be useful for:
- Agricultural question answering
- Crop advisory research
- Evaluation on agriculture-specific benchmarks
- Instruction-following experiments in English and Hindi
The model has not been validated for production deployments or high-stakes agricultural decision-making.
Access
This repository is gated.
Access requests are reviewed on a case-by-case basis. The model is shared solely for research and evaluation purposes. Access may be denied if the intended use does not align with the project's research objectives or licensing terms.
Limitations
- This is a research model and may generate inaccurate, incomplete, or outdated agricultural information.
- Model outputs should not be considered professional agricultural advice.
- Performance may vary across crops, regions, languages, and farming practices.
- Users should independently verify all critical recommendations before taking action.
License
This model is released under the CC BY-NC-SA 4.0 license.
Commercial use, redistribution for commercial purposes, or use in commercial products or services is not permitted without prior authorization from the model authors.