Hercule Telugu LoRA
This Hercule release is the Telugu LoRA-designated member of AI4Bharat's CIA evaluation suite. It derives from Llama-3.1-8B-Instruct.
The evaluator needs a reference and a scoring rubric
Its prompt includes the instruction, candidate response, reference answer and score definitions. It uses English references to judge multilingual answers. The publisher supplies separate language releases and training variants.Hercule Telugu LoRA official model card
Treat the score as one evaluation signal
Practical guidance. Check Telugu judgments against human examples from the domain you need. An English reference can omit a cultural or linguistic detail that matters in the candidate response.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | hercule-te-lora |
| Training data | INTEL; evaluated by the authors on RECON.Hercule Telugu LoRA official model card |
| Output | Written feedback and a score from 1 to 5, using a reference answer and rubric.Hercule Telugu LoRA official model card |
| Terms | The card declares MIT; the Llama 3.1 base has additional upstream terms.Hercule Telugu LoRA official model card |
Related models
- Hercule Telugu
AI4Bharat · Hercule
Fully fine-tuned Llama-3.1-8B-Instruct adaptation for evaluating Telugu answers.
Launch date not confirmedDownloadable - Llama-Prometheus
AI4Bharat · CIA Suite
English answer evaluator adapted from Llama-3.1-8B-Instruct using Feedback-Collection and Prometheus 2 training.
Launch date not confirmedDownloadable