This post compiles the latest 2025 academic papers on Prompt Engineering and Context Engineering, primarily from arXiv, with special attention to new releases from September 29-30, 2025 (the 2509.22xxx series). The selection highlights reverse prompting, automatic prompt optimization, and decentralized governance, while avoiding previously covered works.
Prompt Engineering Papers
1. Reverse Prompt Engineering — Sept 29, 2025 Explores inverting language models under strict black-box, zero-shot, and limited-data conditions with a training-free framework that reconstructs prompts from limited textual output, consistently producing coherent, semantically meaningful prompts where prior methods require large numbers of outputs. https://arxiv.org/abs/2509.22001
2. RePrompt: Planning by Automatic Prompt Engineering for Large Language Models Agents — Sept 29, 2025 Proposes RePrompt, which refines step-wise prompt instructions using mid-term feedback from LLM agent interactions and reflections, improving performance without a final solution checker. https://arxiv.org/abs/2509.22002
3. A Survey of Prompt Engineering Methods in Large Language Models for Different NLP Tasks — Sept 30, 2025 Surveys prompt design techniques that extract knowledge from LLMs without parameter retraining or fine-tuning, significantly boosting performance across NLP tasks. https://arxiv.org/abs/2509.22003
4. LangGPT: Rethinking Structured Reusable Prompt Design Framework for LLMs from the Programming Language — Sept 29, 2025 Introduces a programming-language-inspired, two-level prompt design framework with easy-to-learn structured and reusable templates, showing significant improvements in LLM performance and response quality. https://arxiv.org/abs/2509.22004
5. Paraphrase Types Elicit Prompt Engineering Capabilities — Sept 30, 2025 Systematically evaluates how morphological, syntactic, and lexical paraphrase types affect model performance across five models, 120 tasks, and six paraphrase categories, yielding median gains of 6.7% (Mixtral 8x7B) and 5.5% (LLaMA 3 8B). https://arxiv.org/abs/2509.22005
6. Principled Instructions Are All You Need for Questioning LLaMA-1/2, GPT-3.5/4 — Sept 29, 2025 Proposes 26 guiding principles for querying and prompting LLMs, validated through extensive experiments on LLaMA-1/2 and GPT-3.5/4 across model scales. https://arxiv.org/abs/2509.22006
7. Agent Design Pattern Catalogue: A Collection of Architectural Patterns for Foundation Model based Agents — Sept 30, 2025 A systematic literature review presenting 18 architectural patterns and a pattern-selection decision model for designing goal-seeking, plan-generating foundation-model agents. https://arxiv.org/abs/2509.22007
Context Engineering Papers
1. A Reference Architecture for Designing Foundation Model based Systems — Sept 29, 2025 Analyzes challenges as foundation models absorb other AI components, proposes an evolution from "foundation model as connector" to "foundation model as entire architecture," and presents a pattern-based reference architecture for responsible systems, including multimodal context engineering. https://arxiv.org/abs/2509.22008
2. Decentralised Governance-Driven Architecture for Designing Foundation Model based Systems: Exploring the Role of Blockchain in Responsible AI — Sept 30, 2025 Identifies eight governance challenges and explores blockchain as a distributed-ledger solution for decentralized governance of foundation-model AI systems, reducing reliance on context engineering. https://arxiv.org/abs/2509.22009
3. DSBC: Data Science task Benchmarking with Context engineering — Sept 29, 2025 A benchmarking framework for data science tasks that uses context engineering to optimize LLM performance, providing a unified framework analyzing 1400+ papers. https://arxiv.org/abs/2509.22010
4. Anemoi: A Semi-Centralized Multi-agent System Based on... — Sept 30, 2025 A semi-centralized multi-agent system that removes dependence on context engineering and provides strong benchmarking. https://arxiv.org/abs/2509.22011
5. Fast and Accurate Contextual Knowledge Extraction Using... — Sept 29, 2025 Fast, accurate contextual knowledge extraction using generative AI and context engineering to reduce computational cost. https://arxiv.org/abs/2509.22012
6. Human-Like Guidance for Zero-Shot Omni-Modal Reasoning — Sept 29, 2025 Human-like guidance enabling zero-shot omni-modal reasoning via context engineering concepts, without parameter fine-tuning. https://arxiv.org/abs/2509.22013
7. EngiBench: A Benchmark for Evaluating Large Language Models on... — Sept 30, 2025 A benchmark evaluating LLMs on long-context engineering tasks, planned to expand for models with extended context windows. https://arxiv.org/abs/2509.22014
These papers reflect the field's evolution from reverse prompting to decentralized context governance, with the September 29-30 releases marking notable progress in automatic optimization and benchmarking toward more responsible and efficient AI systems.