Research map / Probabilistic and scientific modeling / Molecules and proteins
TextResNet: Decoupling and Routing Optimization Signals in Compound AI Systems via Deep Residual Tuning
Suizhi Huang, Mei Li, Han Yu, Xiaoxiao Li
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 Additive Semantic Deltas to TextGrad (ablation) | TextGrad · HotpotQA | F1: 24.86 -> 32.15 (+7.29 F1) | Improved |
| Add Additive Semantic Deltas to TextGrad (ablation) | TextGrad · BigCodeBench | pass rate (%): 35.71 -> 36.07 | No clear effect |
| Add Semantic Gradient Decomposition (ablation) | TextGrad + Res · HotpotQA | F1: 32.15 -> 36.69 (+4.54 F1) | Improved |
| Add Semantic Gradient Decomposition (ablation) | TextGrad + Res · BigCodeBench | pass rate (%): 36.07 -> 36.50 | No clear effect |
| Add Causal Routing (ablation) | TextGrad + Res + Proj · HotpotQA | F1: 36.69 -> 37.71 | Improved |
| Add Causal Routing (ablation) | TextGrad + Res + Proj · BigCodeBench | pass rate (%): 36.50 -> 36.70 | No clear effect |
| Add Density-Aware Scheduling (ablation) | TextGrad + Res + Proj + Route · HotpotQA | F1: 37.71 -> 46.23 (+8.52 F1) | Improved |
| Add Density-Aware Scheduling (ablation) | TextGrad + Res + Proj + Route · BigCodeBench | pass rate (%): 36.70 -> 37.86 | Improved |
| Use TextResNet instead of TextGrad | TextGrad · HotpotQA | F1: 24.86 -> 46.23 (+21.37 F1) | Improved |
| Use TextResNet instead of TextGrad | TextGrad · BigCodeBench | pass rate (%): 35.71 -> 37.86 | Improved |
| Use TextResNet instead of DSPy | DSPy · PubMedQA | accuracy: 60.26 -> 60.31 | No clear effect |
| Use TextResNet instead of DSPy | DSPy · STARK-PRIME | MRR: 41.40 -> 41.75 | No clear effect |
Closest papers by meaning
- AgentGrad: Intervention-guided Prompt Optimization for Multi Agent Systems (NeurIPS 2026 · similarity 0.85)
- Revolve: Optimizing AI Systems by Tracking Response Evolution in Textual Optimization (ICML 2025 · similarity 0.85)
- HiVA: Self-organized Hierarchical Variable Agent via Goal-driven Semantic-Topological Evolution (AAAI 2026 · similarity 0.83)
- Meta$^n$: Recursive Self-Improvement through Emergent Depth (NeurIPS 2026 · similarity 0.83)
- EvoOptiGraph: Weakness-Driven Coevolution via Graph-Based Structural Generation for Optimization Modeling (NeurIPS 2026 · similarity 0.82)
- Stable Gradients for Stable Learning at Scale in Deep Reinforcement Learning (NeurIPS 2025 · similarity 0.82)
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.
