dnaihao/qwen2.5-7b-tablellama
dnaihao/qwen2.5-7b-tablellama is a 7.6 billion parameter language model, a replication of TableLlama, fine-tuned from Qwen2.5-7B-Instruct. It is specifically designed and instruction-tuned for tasks involving tabular data, leveraging the tablellama_train.json corpus from dnaihao/Table-Instructs. This model serves as an additional artifact for research on table-focused LLMs, extending experiments from the EACL 2026 Findings paper on model and data effects for table LLMs.
Loading preview...
Overview
dnaihao/qwen2.5-7b-tablellama is a 7.6 billion parameter language model, a replication of the TableLlama architecture. It is fine-tuned from the Qwen2.5-7B-Instruct base model, specifically designed for tasks involving tabular data.
This model was released as an additional artifact alongside the EACL 2026 Findings paper, "What Really Matters for Table LLMs? A Meta-Evaluation of Model and Data Effects" (Deng et al., 2026). It extends the paper's experiments by providing another base-model variant trained on the same specialized corpus, complementing the main research grid which covered Mistral-v0.3, OLMo, and Phi-3-small at the 7B scale.
Key Capabilities
- Table-centric Instruction Following: Specialized in understanding and generating responses based on tabular data, due to its instruction-tuning on the
tablellama_train.jsoncorpus fromdnaihao/Table-Instructs. - Research Artifact: Provides a valuable resource for researchers studying the impact of different base models on table-focused LLM performance.
- Qwen2.5 Foundation: Benefits from the strong base capabilities of the Qwen2.5-7B-Instruct model.
Good for
- Academic Research: Ideal for researchers investigating the effectiveness of various LLM architectures and training data for table-related tasks.
- Benchmarking: Can be used to evaluate and compare performance on table-based benchmarks, adapting the evaluation scripts provided by the creators.
- Developing Table-Aware Applications: Suitable for developers looking to build applications that require robust understanding and interaction with structured tabular data.