TAUR-dev/rankalign-v6-gemma-2-2b-d0.15-e2-hc-b2d-dbl-all-tco-ln-nv1-ng1-vlo-fsx

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

TAUR-dev/rankalign-v6-gemma-2-2b-d0.15-e2-hc-b2d-dbl-all-tco-ln-nv1-ng1-vlo-fsx is a 2.6 billion parameter model fine-tuned from Google's Gemma-2-2b base model. This model is part of the rankalign project, specifically optimized for tasks involving hypernym-concat-bananas-to-dogs-double-all, utilizing online typicality correction and length normalization. It is designed for specific linguistic evaluation tasks, particularly those related to hypernym identification and validation.

Loading preview...

Model Overview

This model, rankalign-v6-gemma-2-2b-d0.15-e2-hc-b2d-dbl-all-tco-ln-nv1-ng1-vlo-fsx, is a fine-tuned checkpoint derived from the Google Gemma-2-2b base model. It is developed as part of the rankalign project, focusing on specialized linguistic tasks.

Training Details

The model underwent specific fine-tuning with the following key parameters:

  • Base Model: google/gemma-2-2b
  • Version: v6
  • Task: hypernym-concat-bananas-to-dogs-double-all
  • Epochs: 2
  • Delta: 0.15
  • Typicality Correction: Online
  • Length Normalization: Enabled
  • Preference Loss Weight: 1
  • NLL Validator Weight: 1
  • NLL Generator Weight: 1
  • Validator Log-Odds: Enabled
  • Force Same-X: Enabled

Key Capabilities

This model is specifically configured for:

  • Hypernym Identification: Designed to evaluate and process hypernym-related tasks, as indicated by its training task.
  • Linguistic Evaluation: Optimized for specific evaluation scripts that assess its performance on various hypernym datasets (e.g., hypernym-bananas, hypernym-dogs).
  • Controlled Generation: Incorporates features like typicality correction and length normalization, suggesting a focus on generating more controlled and relevant outputs for its target tasks.

When to Use This Model

This model is particularly suited for researchers and developers working on:

  • Linguistic Analysis: Specifically for tasks involving hypernym relationships and semantic hierarchies.
  • Evaluation of Language Models: As a tool to evaluate the performance of language models on fine-grained linguistic understanding, particularly in the context of hypernyms.
  • Reproducing Rankalign Research: For those looking to reproduce or build upon the methodologies presented in the rankalign project.

Finetunes
1 models