soramikaduki/StarAntler-RP-WestLake-chatvector

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kLicense:apache-2.0Architecture:Transformer0.0K Open Weights Cold

The soramikaduki/StarAntler-RP-WestLake-chatvector is a 7B parameter language model based on the Mistral architecture, created by soramikaduki through a DARE TIES merge. This model specifically enhances NSFW capabilities by combining two ChatVector-enhanced models, Aratako/Antler-7B-RP and Japanese-Starling-ChatV-7B-RP, with senseable/WestLake-7B-v2. It is designed for roleplay scenarios requiring robust NSFW content generation, achieving an ELYZA-tasks-100 average score of 3.16.

Loading preview...

Model Overview

The soramikaduki/StarAntler-RP-WestLake-chatvector is a 7B parameter language model, developed by soramikaduki, that leverages the Mistral architecture. It was created using the DARE TIES merge method, combining two distinct models that were themselves enhanced using the ChatVector technique.

Key Capabilities and Enhancements

This model's primary focus is on strengthening NSFW (Not Safe For Work) content generation, particularly within roleplay contexts. It achieves this by merging:

  • A ChatVector-enhanced version of Aratako/Antler-7B-RP with senseable/WestLake-7B-v2.
  • A ChatVector-enhanced version of Aratako/Japanese-Starling-ChatV-7B-RP with senseable/WestLake-7B-v2.

The merge process assigned equal weights to these two enhanced models, resulting in a specialized model for explicit roleplay scenarios.

Performance

On the ELYZA-tasks-100 benchmark, which evaluates instruction-tuned Japanese models, this model achieved an average score of 3.16. This score was assessed by gpt-4-0125-preview.