GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
AxisMeru
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
AxisMeru
Model Tree
AxisMeru/prabhasa-nyaya-element-judge-4b-arm0
Input Modalities
Output Modalities
Supported Functionality
A LoRA adapter (rank 64) trained on top of the production 4B element judge AxisMeru/prabhasa-nyaya-element-judge-4b-arm0 (weights file sha256 c2fa5954d5bab3885d13d472e38aad67fe6cbad51f54a08a472fa2f9a6a9eec6), on a very small set of real-text element examples extracted by two model readers (claude-sonnet-5) with no human review. Reader agreement is not accuracy.
Status: a training candidate for a pre-registered evaluation. It is NOT the production judge, it is not adopted, and nothing here is a performance claim; any claim will be made only through the pre-registered evaluation, labelled "dev stack (local)". The training recipe was amended after the training-set size was known (recorded in the project's seal record).
Credit: some of the training examples are built from the Indian High Court Judgments data set (Dattam Labs, licensed CC BY 4.0, https://registry.opendata.aws/indian-high-court-judgments), selected and annotated by us.
Files: adapter_model.safetensors (sha256 cdfb36de5eab4f3dc0b7e9daa16698ef813e1f80eb377410949588e54d05f8b3) and adapter_config.json (base model path rewritten to the public repo id; the weights file is byte-identical to the one trained). No training data, quotes or identifiers are included.