myyycroft/Gemma-4-E2B-AmbigQA-full-short-form-prompt-member-4

VISIONConcurrent Unit Cost:1Model Size:5.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 7, 2026Architecture:Transformer Featherless Exclusive Cold

myyycroft/Gemma-4-E2B-AmbigQA-full-short-form-prompt-member-4 is a 5.1 billion parameter Gemma-4-E2B-it model fine-tuned by myyycroft specifically for open-domain factoid question answering. This model is an ensemble member optimized to provide short-form answers, typically 1-4 words, without explanations or preambles. It excels at extracting precise answers like names, places, dates, or numbers directly from questions, making it suitable for applications requiring concise factual responses.

Loading preview...

Model Overview

This model, myyycroft/Gemma-4-E2B-AmbigQA-full-short-form-prompt-member-4, is an ensemble member (specifically, member 4 with seed 4078) derived from the google/gemma-4-E2B-it architecture. It has been fine-tuned on the AmbigQA dataset to specialize in open-domain factoid question answering, focusing on generating short-form answers.

Key Capabilities

  • Short-form Factoid Answering: Designed to provide concise answers (1-4 words) such as names, places, dates, or numbers, without additional context or explanation.
  • Strict Formatting: Adheres to a strict output format, providing only the answer string on a single line, without restating the question, markdown, or quotation marks.
  • AmbigQA Optimization: Fine-tuned on the sewon/ambig_qa dataset, indicating a focus on handling potentially ambiguous questions by providing a canonical short form.

Performance Insights

Evaluation metrics are based on small, fixed subsets (AmbigQA: 128, IFEval: 64, MMLU: 228) at the final training step (1620 steps, 3 epochs). This specific member achieved:

  • AmbigQA (128) accuracy: 0.1016
  • AmbigQA (128) AlignScore: 0.1764
  • IFEval (64) prompt_level_strict_accuracy: 0.6719
  • MMLU (228) accuracy: 0.5439

Use Cases

This model is particularly well-suited for applications where direct, unambiguous, and short factual answers are required. Examples include:

  • Chatbots: Providing quick answers to user queries.
  • Knowledge Retrieval: Extracting specific data points from text.
  • Automated Q&A Systems: Generating concise responses for fact-based questions.

Finetunes
1 models