# Visual Editing with LLM-based Tool Chaining: An Efficient Distillation Approach for Real-Time Applications

arXiv: https://arxiv.org/abs/2410.02952 · HTML: https://mlsanity.com/paper/2410.02952/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 28 references that are in the map.
- First posted on arXiv 3 Oct 2024.

## Builds on (most cited first, 23 of 23)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Toolformer: Language Models Can Teach Themselves to Use Tools](https://mlsanity.com/paper/2302.04761/) (2023)
- [LAION-5B: An open large-scale dataset for training next generation image-text models](https://mlsanity.com/paper/2210.08402/) (2022)
- [Scaling Instruction-Finetuned Language Models](https://mlsanity.com/paper/2210.11416/) (2022)
- [Sequence to Sequence Learning with Neural Networks](https://mlsanity.com/paper/1409.3215/) (2014)
- [Gorilla: Large Language Model Connected with Massive APIs](https://mlsanity.com/paper/2305.15334/) (2023)
- [Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes](https://mlsanity.com/paper/2305.02301/) (2023)
- [Visual ChatGPT: Talking, Drawing and Editing with Visual Foundation Models](https://mlsanity.com/paper/2303.04671/) (2023)
- [MM-REACT: Prompting ChatGPT for Multimodal Reasoning and Action](https://mlsanity.com/paper/2303.11381/) (2023)
- [FateZero: Fusing Attentions for Zero-shot Text-based Video Editing](https://mlsanity.com/paper/2303.09535/) (2023)
- [TinyStories: How Small Can Language Models Be and Still Speak Coherent English?](https://mlsanity.com/paper/2305.07759/) (2023)
- [TokenFlow: Consistent Diffusion Features for Consistent Video Editing](https://mlsanity.com/paper/2307.10373/) (2023)
- [Pix2Video: Video Editing using Image Diffusion](https://mlsanity.com/paper/2303.12688/) (2023)
- [GPT4Tools: Teaching Large Language Model to Use Tools via Self-instruction](https://mlsanity.com/paper/2305.18752/) (2023)
- [LLaVA-Plus: Learning to Use Tools for Creating Multimodal Agents](https://mlsanity.com/paper/2311.05437/) (2023)
- [RAVE: Randomized Noise Shuffling for Fast and Consistent Video Editing with Diffusion Models](https://mlsanity.com/paper/2312.04524/) (2023)
- [LLM2LLM: Boosting LLMs with Novel Iterative Data Enhancement](https://mlsanity.com/paper/2403.15042/) (2024)
- [VidEdit: Zero-Shot and Spatially Aware Text-Driven Video Editing](https://mlsanity.com/paper/2306.08707/) (2023)
- [RecAI: Leveraging Large Language Models for Next-Generation Recommender Systems](https://mlsanity.com/paper/2403.06465/) (2024)
- [TACT: Advancing Complex Aggregative Reasoning with Information Extraction Tools](https://mlsanity.com/paper/2406.03618/) (2024)
- [Adjusting Image Attributes of Localized Regions with Low-level Dialogue](https://mlsanity.com/paper/2002.04678/) (2020)
- [CAISE: Conversational Agent for Image Search and Editing](https://mlsanity.com/paper/2202.11847/) (2022)

## Abstract (arXiv)

We present a practical distillation approach to fine-tune LLMs for invoking tools in real-time applications. We focus on visual editing tasks; specifically, we modify images and videos by interpreting user stylistic requests, specified in natural language ("golden hour"), using an LLM to select the appropriate tools and their parameters to achieve the desired visual effect. We found that proprietary LLMs such as GPT-3.5-Turbo show potential in this task, but their high cost and latency make them unsuitable for real-time applications. In our approach, we fine-tune a (smaller) student LLM with guidance from a (larger) teacher LLM and behavioral signals. We introduce offline metrics to evaluate student LLMs. Both online and offline experiments show that our student models manage to match the performance of our teacher model (GPT-3.5-Turbo), significantly reducing costs and latency. Lastly, we show that fine-tuning was improved by 25% in low-data regimes using augmentation.