How many instructions can LLMs follow at once?

Subscribers:
6,300
Published on ● Video Link: https://www.youtube.com/watch?v=B_MuR6NFga4



Duration: 0:00
490 views
29


As LLMs are able to do more complex tasks, how many instructions should we give it at one go for reliable, robust generation?

This ability for LLM to follow a greater number of instructions will help it greatly in doing more complex tool use / multi-step reasoning / agentic tasks.

By constraining LLMs to include difficult financial keywords in the report generated, this study measures how 20 SOTA LLMs can handle increasing constraints.

Overall, it appears that o3 (reasoning) and gemini-2.5-pro-review (reasoning) are the best at following complex instructions.

On another note, should we aim to increase instruction following complexity, or should we aim to modularise the process into easy bite-sized bits?

~~~

Slides: https://github.com/tanchongmin/john-youtube/blob/main/Discussion_Sessions/LLM_Instructions.pdf
Paper: https://arxiv.org/pdf/2507.11538

Other references:
T5 Paper: https://arxiv.org/pdf/1910.10683
Length and correctness in LLMs (longer response tends to be inaccurate): https://arxiv.org/html/2505.00127v1

My repositories mentioned:
StrictJSON: https://github.com/tanchongmin/strictjson
AgentJo: https://github.com/tanchongmin/agentjo
text-rpg (my attempt at vibe-coding an RPG): https://github.com/tanchongmin/text-rpg
Between Underthinking and Overthinking: An Empirical Study of Reasoning

~~~

0:00 Introduction
5:21 Main Results
16:22 Why is instruction following important?
24:37 Experiment Details
30:21 Report Generation Prompt
42:22 Verbosity of Response vs Accuracy
47:06 Variability of Accuracy across models
1:00:41 Does reasoning help with instruction following?
1:12:22 My guidelines: How to use LLMs in a process / agentic flow
1:28:16 Discussion
1:38:20 Conclusion

~~~

AI and ML enthusiast. Likes to think about the essences behind breakthroughs of AI and explain it in a simple and relatable way. Also, I am an avid game creator.

Discord: https://discord.gg/bzp87AHJy5
LinkedIn: https://www.linkedin.com/in/chong-min-tan-94652288/
Online AI blog: https://delvingintotech.wordpress.com/
Twitter: https://twitter.com/johntanchongmin
Try out my games here: https://simmer.io/@chongmin




Other Videos By John Tan Chong Min


2025-09-08DINOv3: One backbone, multiple image/video tasks
2025-08-18R-Zero: Self-Evolving Reasoning LLM from Zero Data
2025-08-11Reasoning without Language (Part 2) - Deep Dive into 27 mil parameter Hierarchical Reasoning Model
2025-08-04Reasoning without Language - Deep Dive into 27 mil parameter Hierarchical Reasoning Model
2025-07-28No need for symbolic programs for Math? Natural language approach to IMO
2025-07-21How many instructions can LLMs follow at once?
2025-07-15Arjo Chakravarty: Indoor Localisation with Visual Language Models (VLMs)
2025-07-14MemOS: A Paradigm Shift to Memory as a First Class Citizen for LLMs
2025-07-07Multimodal Query for Images: Text/Image Multimodal Query with Negative Filter and Folder Selection
2025-06-30Universal Filter (Part 4 - Finale): Knowledge/Memory, Reflection, Communication between Individuals
2025-06-23Universal Filter (Part 3): Learning the Filters, Universal Database, Individual Knowledge Base
2025-06-16Universal Filter (Part 2): Time, Akashic Records, Individual Mind-based, Body-based memory
2025-06-04Good Vibes Only with Dylan Chia: Lyria (Music), Veo3 (Video), Gamma (Slides), GitHub Copilot (Code)
2025-03-10Memory Meets Psychology - Claude Plays Pokemon: How It works, How to improve it
2025-02-24Vibe Coding: How to use LLM prompts to code effectively!
2025-01-26PhD Thesis Overview (Part 2): LLMs for ARC-AGI, Task-Based Memory-Infused Learning, Plan for AgentJo
2025-01-20PhD Thesis Overview (Part 1): Reward is not enough; Towards Goal-Directed, Memory-based Learning
2024-12-04AgentJo CV Generator: Generate your CV by searching for your profile on the web!
2024-11-11Can LLMs be used in self-driving? CoMAL: Collaborative Multi-Agent LLM for Mixed Autonomy Traffic
2024-10-28From TaskGen to AgentJo: Creating My Life Dream of Fast Learning and Adaptable Agents
2024-10-21Tian Yu X John: Discussing Practical Gen AI Tips for Image Prompting