Research map / Large language models / Agents, code and math
LLM2Fx-Tools: Tool Calling for Music Post-Production
SeungHeon Doh, Junghyun Koo, Marco A. Martínez-Ramírez, Woosung Choi, Wei-Hsiang Liao, Qiyu Wu, Juhan Nam, Yuki Mitsufuji
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add chain-of-thought planning (ablation) | LLM2Fx-Tools · LP-Fx test (reverse engineering) | module classification accuracy: 67% -> 80% | Improved |
| add chain-of-thought planning (ablation) | LLM2Fx-Tools · LP-Fx test (reverse engineering) | order correlation: 0.49 -> 0.56 | Improved |
| add Number Token Loss (ablation) | LLM2Fx-Tools · LP-Fx test (reverse engineering) | parameter MAE: 0.32 -> 0.23 | Improved |
| add Number Token Loss (ablation) | LLM2Fx-Tools · LP-Fx test (reverse engineering) | MRS L/R distance: 3.69 -> 3.13 | Improved |
| add multi-stage training with adapter pretraining (ablation) | LLM2Fx-Tools · LP-Fx test (reverse engineering) | module classification accuracy: 76% -> 80% | Improved |
| add multi-stage training (ablation) | LLM2Fx-Tools · LP-Fx test (reverse engineering) | AFx-Rep similarity: 0.67 -> 0.68 | Improved |
| replace regression with tool-calling LLM with module selection and ordering | LLM2Fx-Tools vs Regression · LP-Fx test (reverse engineering) | classification accuracy: 55% -> 80% | Improved |
| use regression head instead of LLM for parameters | Regression · LP-Fx test (reverse engineering) | parameter MAE: 0.20 vs 0.23 | Mixed |
| add effect ordering (LLM2Fx-Tools vs MultiTask) at similar DSP distance | LLM2Fx-Tools · LP-Fx test (reverse engineering) | MRS L/R distance: 3.17 -> 3.13 | Improved |
| use LLM2Fx-Tools instead of Gemini 2.5 Flash | LLM2Fx-Tools · Audio effects style transfer (MoisesDB to MedleyDB) | AF distance: 9.00 -> 7.41 | Improved |
| use differentiable DeepAFx-ST instead of LLM2Fx-Tools | DeepAFx-ST · Audio effects style transfer (MoisesDB to MedleyDB) | AF distance: 10.50 vs 7.41 | Hurt |
| use LLM2Fx-Tools in listening test vs Gemini 2.5 Flash | LLM2Fx-Tools · MUSHRA listening test | MUSHRA score: 62.8 vs 56.5 | Improved |
Closest papers by meaning
- LLark: A Multimodal Instruction-Following Language Model for Music (ICML 2024 · similarity 0.87)
- FilmComposer: LLM-Driven Music Production for Silent Film Clips (CVPR 2025 · similarity 0.86)
- ThinkSound: Chain-of-Thought Reasoning in Multimodal LLMs for Audio Generation and Editing (NeurIPS 2025 · similarity 0.86)
- CMI-RewardBench: Evaluating Music Reward Models with Compositional Multimodal Instruction (ICML 2026 · similarity 0.84)
- Text2midi: Generating Symbolic Music from Captions (AAAI 2025 · similarity 0.84)
- MID-FiLD: MIDI Dataset for Fine-Level Dynamics (AAAI 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
