LlamaIndex Webinar: Advanced Tabular Data Understanding with LLMs

Using LLMs to do question-answering and understanding over tabular data is challenging. The naive approaches (dump text into prompt, text-to-SQL), oftentimes don't work well over tabular data.
​In this webinar we're excited to feature the authors of two papers featuring advanced techniques for tabular data understanding with LLMs:
1. ​Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understanding, featuring Zilong Wang: arxiv.org/abs/2401.04398v1
2. ​Rethinking Tabular Data Understanding with Large Language Models, featuring Tianyang Liu: arxiv.org/abs/2312.16702
​Both papers pinpoint shortcomings with naive approaches and propose novel techniques to reason about tabular data in a robust manner. Come check it out!
​We've also implemented these as LlamaPack templates:
1. ​Chain-of-Table LlamaPack: llamahub.ai/l/llama_packs-tab...
2. ​Mix-Self-Consistency LlamaPack: llamahub.ai/l/llama_packs-tab...
Timeline:
00:00-22:45 - Chain-of-Table
22:45-24:23 - Short Break (skip this part)
24:23-45:00 - Rethinking Tabular Data Understanding + Mix-Self-Consistency
45:00 - Q&A

[@quickcinemarecap]

00:02 Advanced tabular data understanding with LLMs
02:24 LLMs are trained to understand and perform tabular reasoning
07:56 Understanding the operation history for tabular data processing
10:46 Tabular data understanding and experiment overview
16:26 Tabular data is divided into small, medium, and large categories based on token numbers
18:51 Using operations during prompting to make large tables smaller and improve prompt reasoning ability.
24:17 Exploring large language models' understanding of tabular data
26:38 Language models struggle with handling tabular data due to linearizing inputs.
31:17 Table Transportation Impact on Large Language Models
33:31 Importance of headings located in the first column
38:04 Textual reasoning challenges in tabular data understanding
40:37 Texture reasoning provides more robust reasoning than symbolic reasoning
45:16 As LLMs advance, the need for handcrafted techniques for tabular data will evolve over time.
47:43 Tabular data reasoning will benefit from evolving models.
52:21 Advanced techniques for handling tabular data.

[@lukeskywalker7029]

Unfortunately the Chain-of_tables pack link leads to a 404 ...

[@ragsAI]

🚀🚀

[@huyvo9105]

which is better :vv

[@tommyjado127]

so so so so so so so so...

[@lizhong5677]

Very insightful work!!