Abhishek0219/qwen-fine-tuned

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 6, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

The Abhishek0219/qwen-fine-tuned model is a 4 billion parameter Qwen3-4B-Instruct variant, fine-tuned by Abhishek0219 for classifying query-product relevance. It specializes in categorizing Amazon shopping queries into Exact, Substitute, Complement, or Irrelevant, achieving a micro-F1 score of 0.642 on the ESCI test set. This model is optimized for single-token classification tasks, outperforming gpt-4o-mini in zero-shot and few-shot scenarios for this specific relevance classification.

Loading preview...

Model Overview

This model, Abhishek0219/qwen-fine-tuned, is a 4 billion parameter variant of Qwen/Qwen3-4B-Instruct-2507 that has been specifically fine-tuned for product query relevance classification. It leverages QLoRA (r=16, alpha=32) on Amazon's Shopping Queries Dataset (ESCI, KDD Cup 2022) to determine if a query-product pair is Exact, Substitute, Complement, or Irrelevant.

Key Capabilities

  • Specialized Classification: Designed for precise, single-token classification of query-product relevance.
  • Performance: Achieves a micro-F1 score of 0.642 on the 5k-row ESCI test set, surpassing gpt-4o-mini's 0-shot (0.612) and 4-shot (0.543) performance on this specific task.
  • Efficient Output: Optimized to return exactly one token, allowing for direct interpretation of logprobs for label and confidence.

Good For

  • E-commerce Applications: Ideal for systems requiring automated classification of customer shopping queries against product listings.
  • Relevance Ranking: Can be integrated into search engines or recommendation systems to improve product relevance.
  • Resource-Constrained Environments: As a 4B parameter model, it offers a balance of performance and computational efficiency for specialized tasks.