dnaihao/phi-3-mini-tablegpt

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:4kPublished:Aug 1, 2025License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

dnaihao/phi-3-mini-tablegpt is a 4-billion parameter language model, based on Microsoft's Phi-3-mini-4k-instruct, specifically fine-tuned for table-related tasks. It replicates the TableGPT architecture and was trained on the Table-Instructs corpus, offering specialized capabilities for understanding and generating content from structured tabular data. This model is an artifact from the EACL 2026 Findings paper, extending research into model and data effects for Table LLMs, making it ideal for applications requiring robust table processing.

Loading preview...

Overview

dnaihao/phi-3-mini-tablegpt is a 4-billion parameter language model, a replication of the TableGPT architecture. It is built upon Microsoft's Phi-3-mini-4k-instruct base model and was fine-tuned using the tablegpt_large_train.json corpus from the dnaihao/Table-Instructs dataset. This model serves as an additional artifact for the EACL 2026 Findings paper, "What Really Matters for Table LLMs? A Meta-Evaluation of Model and Data Effects," contributing to research on model and data effects in table-focused Large Language Models.

Key Capabilities

  • Specialized Table Understanding: Fine-tuned specifically for tasks involving structured tabular data, enabling robust interpretation and generation.
  • Research Replication: Provides a direct replication of the TableGPT model, allowing for further research and experimentation in table-centric AI.
  • Phi-3-mini Foundation: Benefits from the compact yet capable architecture of the Phi-3-mini-4k-instruct base model, offering a balance of performance and efficiency.

Good For

  • Table-based Question Answering: Ideal for applications that require extracting information or answering questions based on structured tables.
  • Data Analysis and Summarization: Useful for generating summaries or insights from tabular datasets.
  • Academic Research: Serves as a valuable tool for researchers exploring the performance and characteristics of LLMs on table-related tasks, particularly in the context of the EACL 2026 paper.

Finetunes
1 models