TAUR-dev/rankalign-v6-gemma-2-2b-d0.15-e2-hc-b2d-dbl-all-tco-vlo-fsx

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:2.6BQuant:BF16Ctx Length:8kPublished:Apr 6, 2026Architecture:Transformer Warm

The TAUR-dev/rankalign-v6-gemma-2-2b-d0.15-e2-hc-b2d-dbl-all-tco-vlo-fsx model is a 2.6 billion parameter language model fine-tuned from Google's Gemma-2-2b base model. This specific version, v6, is optimized for hypernym-concat-bananas-to-dogs-double-all tasks, focusing on identifying hierarchical relationships between concepts. It utilizes an online typicality correction and validator log-odds during its training process. The model is designed for specialized tasks involving semantic hierarchy and relation extraction within a context length of 8192 tokens.

Loading preview...

Model Overview

TAUR-dev/rankalign-v6-gemma-2-2b-d0.15-e2-hc-b2d-dbl-all-tco-vlo-fsx is a 2.6 billion parameter language model derived from the google/gemma-2-2b base architecture. It represents version 6 of the rankalign project's fine-tuned checkpoints, specifically trained for tasks involving hypernym identification and concatenation.

Key Training Details

This model underwent a specialized training regimen focused on the hypernym-concat-bananas-to-dogs-double-all task over two epochs. Notable training parameters include:

  • Base Model: google/gemma-2-2b
  • Task Focus: Hypernym concatenation, specifically from 'bananas' to 'dogs' with a 'double-all' configuration.
  • Delta: 0.15
  • Epochs: 2
  • Typicality Correction: Applied online during training.
  • Validator Log-Odds: Enabled, indicating a focus on validating logical consistency in predictions.
  • Force Same-X: True, suggesting a constraint on input processing.

Use Cases

This model is particularly suited for research and applications requiring precise identification and generation of hypernymic relationships. Its fine-tuning on specific hypernym tasks suggests strong performance in:

  • Semantic Hierarchy Extraction: Identifying 'is-a' relationships between words or concepts.
  • Knowledge Graph Construction: Populating or validating hierarchical structures in knowledge bases.
  • Specialized NLP Tasks: Any application where understanding and generating hypernyms is critical, as demonstrated by the extensive evaluation scripts provided for various hypernym tasks (e.g., hypernym-bananas, hypernym-dogs, hypernym-elephants).

Developers can evaluate the model's performance on these specific tasks using the provided scripts/eval.py commands, which include parameters for typicality correction and validator log-odds.