Research map / Large language models / Language, safety and interpretability
Beyond Magic Words: Sharpness-Aware Prompt Evolving for Robust Large Language Models with TARE
Guancheng Wan, Lucheng Fu, Haoxin Liu, Yiqiao Jin, Hui Yi Leong, Eric Hanchen Jiang, Hejia Geng, Jinhe Bi, Yunpu Ma, Xiangru Tang, B. Aditya Prakash, Yizhou Sun and 1 more
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 anisotropic adaptive search (ATARE over TARE) | TARE, Llama 3.1 8B Instruct · Tracking Shuffled Objects (GPT-4o) | accuracy (%): 90.0 -> 93.5 | Improved |
| add anisotropic adaptive search (ATARE over TARE) | TARE, Llama 3 8B Instruct · Object Counting (Claude 3.5 Sonnet) | accuracy (%): 89.5 -> 93.6 | Improved |
| add anisotropic adaptive search (ATARE over TARE) | TARE, Llama 3.1 8B Instruct · GSM8K (GPT-4o) | accuracy (%): 86.2 -> 86.4 | No clear effect |
| use TARE instead of TextGrad | TextGrad, Llama 3.1 8B Instruct · Object Counting (GPT-4o) | accuracy (%): 88.6 -> 91.0 | Improved |
| use ATARE instead of TextGrad | TextGrad, Llama 3.1 8B Instruct · Tracking Shuffled Objects (GPT-4o) | accuracy (%): 86.3 -> 93.5 | Improved |
| use ATARE instead of TextGrad | TextGrad, GPT-4o · GPQA | accuracy (%): 52.5 -> 53.0 | Improved |
| use ATARE instead of TextGrad | TextGrad, Llama 3.1 8B Instruct · GPQA | accuracy (%): 30.8 -> 33.8 | Improved |
| use ATARE instead of TextGrad | TextGrad, Qwen 2.5 7B Instruct · MMLU College Physics | accuracy (%): 78.4 -> 79.4 | Improved |
| use ATARE instead of TextGrad | TextGrad, Llama 3 8B Instruct · GSM8K (GPT-4o) | accuracy (%): 81.8 -> 81.8 | No clear effect |
| remove inner adversarial search, LATO or robust validation (ablation) | TARE/ATARE, Llama 3.1 8B Instruct · ablation (figure) | accuracy (%) | Improved |
| replace Generator or Optimizer oracle with Llama 3.1 8B (ablation) | TARE/ATARE · BBH tasks | accuracy (%): within 5% margin | No clear effect |
| reduce search budgets K_t and K~ from 3 to 1 or 2 (ablation) | TARE/ATARE · BBH tasks | accuracy (%) | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Revolve — Revolve: Optimizing AI Systems by Tracking Response Evolution in Textual Optimization (ICML 2025) · better on 8 of 8 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Evolving Prompts In-Context: An Open-ended, Self-replicating Perspective (ICML 2025 · similarity 0.89)
- On the Worst Prompt Performance of Large Language Models (NeurIPS 2024 · similarity 0.89)
- Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting (ICLR 2024 · similarity 0.88)
- Scaling Prompt Synthesis for Large Language Model Reasoning (ICML 2026 · similarity 0.88)
- Brittlebench: Quantifying LLM robustness via prompt sensitivity (NeurIPS 2026 · similarity 0.88)
- Unleashing the Potential of Large Language Models as Prompt Optimizers: Analogical Analysis with Gradient-based Model Optimizers (AAAI 2025 · similarity 0.88)
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.
