LLM-AutoDiff: Auto-Differentiate Any LLM Workflow
About
Large Language Models (LLMs) have reshaped natural language processing, powering applications from multi-hop retrieval and question answering to autonomous agent workflows. Yet, prompt engineering -- the task of crafting textual inputs to effectively direct LLMs -- remains difficult and labor-intensive, particularly for complex pipelines that combine multiple LLM calls with functional operations like retrieval and data formatting. We introduce LLM-AutoDiff: a novel framework for Automatic Prompt Engineering (APE) that extends textual gradient-based methods (such as Text-Grad) to multi-component, potentially cyclic LLM architectures. Implemented within the AdalFlow library, LLM-AutoDiff treats each textual input as a trainable parameter and uses a frozen backward engine LLM to generate feedback-akin to textual gradients -- that guide iterative prompt updates. Unlike prior single-node approaches, LLM-AutoDiff inherently accommodates functional nodes, preserves time-sequential behavior in repeated calls (e.g., multi-hop loops), and combats the "lost-in-the-middle" problem by isolating distinct sub-prompts (instructions, formats, or few-shot examples). It further boosts training efficiency by focusing on error-prone samples through selective gradient computation. Across diverse tasks, including single-step classification, multi-hop retrieval-based QA, and agent-driven pipelines, LLM-AutoDiff consistently outperforms existing textual gradient baselines in both accuracy and training cost. By unifying prompt optimization through a graph-centric lens, LLM-AutoDiff offers a powerful new paradigm for scaling and automating LLM workflows - mirroring the transformative role that automatic differentiation libraries have long played in neural network research.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Multi-hop Question Answering | HotpotQA (test) | -- | 334 | |
| Sentiment Analysis | SST-5 (test) | Accuracy55.7 | 189 | |
| Question Classification | TREC (test) | Accuracy86.67 | 150 | |
| Text Classification | TREC (test) | Accuracy85 | 132 | |
| Science Question Answering | ARC Challenge (test) | Accuracy91.1 | 60 | |
| Math Word Problem Solving | GSM8K official 1.3k set (test) | Accuracy88.7 | 53 | |
| Mathematical Problem Solving | MATH (test) | Accuracy89.72 | 45 | |
| Instruction Following | IFBench (test) | Score38 | 36 | |
| Math | GSM8K (test) | Mean@489 | 28 | |
| Prompt Optimization | Prompt Optimization Benchmark | Accuracy56.8 | 24 |