185 articles tagged with #prompt-engineering. AI-curated summaries with sentiment analysis and key takeaways from 50+ sources.
Researchers demonstrate that execution-based voting methods for LLM code generation significantly outperform text-based majority voting by 18-52 percentage points. The study reveals that input quality—particularly sketch-based generation—matters far more than the aggregation algorithm itself, challenging assumptions about how to select optimal code outputs.
Researchers introduce QD-LLM, a framework that evolves lightweight prompt embeddings (~32K parameters) to steer frozen large language models toward diverse outputs without fine-tuning. The approach outperforms existing quality-diversity optimization methods by 46.4% in coverage and demonstrates practical applications in test generation and training data improvement.
Researchers introduce SCALAR, an Actor-Critic-Judge framework that systematically evaluates how AI agents improve through human feedback on theoretical physics problems. The study reveals that multi-turn dialogue consistently outperforms single attempts, but the effectiveness of different feedback strategies depends heavily on the specific pairing of AI models used, with asymmetric model pairs benefiting most from structured critique.
Researchers conducted a controlled empirical study evaluating three LLMs (Claude Haiku, DeepSeek-Chat, Gemini 2.5 Flash) for qualitative coding of psychological safety in software engineering communities. Multi-shot prompting improved Claude Haiku's performance but not the others, while all models exhibited systematic biases in coding predictions, providing evidence-based guidelines for LLM-assisted qualitative research.
Researchers introduce CyBiasBench, a benchmark revealing that LLM agents deployed for cybersecurity attacks exhibit inherent biases toward specific attack families regardless of prompting. The study demonstrates agents resist steering away from their preferred attack patterns, suggesting these biases are fundamental agent characteristics rather than prompt-dependent behaviors.
Researchers introduce CA-SQL, an advanced Text-to-SQL pipeline that dynamically allocates computational resources based on task complexity to improve LLM reasoning. The method achieves state-of-the-art performance on the BIRD benchmark's challenging tier using only GPT-4o-mini, outperforming larger models and demonstrating the efficiency gains possible through intelligent inference-time optimization.
Researchers present CWE-BENCH-PYTHON, a large-scale benchmark demonstrating that poorly formulated prompts significantly increase the likelihood of LLMs generating insecure code. The study shows advanced prompting techniques like Chain-of-Thought can effectively mitigate these security risks, establishing prompt quality as a critical factor in AI-generated code safety.
Prober.ai is an LLM-powered web-based writing environment that uses constrained AI personas and gated feedback mechanisms to improve argumentative writing through inquiry-based questioning rather than text generation. The system addresses cognitive outsourcing in education by forcing student reflection before revealing revision suggestions, grounded in Toulmin's argumentation theory and peer feedback research.
Taklif.AI is an LLM-powered educational platform that generates personalized college assignments based on students' interests and cultural contexts rather than just academic performance metrics. The system uses Llama 3.3 70B with AWS serverless architecture and achieved 84% positive reception in preliminary testing with 68 participants.
Researchers have developed a visual fingerprinting method to compare Large Language Model outputs across different generation conditions by analyzing linguistic choices in content, expression, and structure. This approach enables pattern recognition in LLM behavior that is difficult to detect through individual responses or standard metrics, advancing model evaluation and prompt optimization techniques.
Researchers introduce MASPO, a framework that automatically optimizes prompts across multi-agent LLM systems by evaluating how well each agent's outputs enable downstream success rather than in isolation. The approach uses evolutionary beam search to navigate prompt spaces and achieves 2.9% average accuracy improvements over existing methods across six diverse tasks.
Researchers propose 'mise en place' (MEP), a three-phase preparation methodology for AI coding agents that emphasizes contextual grounding, collaborative specification, and task decomposition before implementation. The approach counters prevalent 'vibe coding' practices by demonstrating that deliberate preparation reduces debugging overhead and enables efficient parallel agent execution, validated through a hackathon case study.
Researchers introduce Memory Inception (MI), a training-free method for steering large language models by inserting text-derived key-value banks at selected attention layers rather than caching full prompts. MI achieves competitive control with instruction prompting while using up to 118x less storage and outperforms existing activation steering methods on personality, reasoning, and guidance tasks.
RaguTeam won SemEval-2026 Task 8 using a seven-model LLM ensemble with a GPT-4o-mini judge selector, achieving a conditioned harmonic mean of 0.7827 and significantly outperforming the baseline. The research demonstrates that model diversity across families, scales, and prompting strategies drives superior performance in multi-turn response generation tasks.
Researchers introduce SafeRedir, an inference-time framework that safely redirects unsafe prompts in image generation models by rerouting them toward benign semantic regions without modifying underlying model weights. The lightweight approach uses token-level embedding interventions to mitigate generation of NSFW content and copyrighted styles while maintaining image quality and resisting adversarial attacks.
Researchers propose NDBench, a benchmark framework testing how frontier LLMs adapt outputs when given neurodivergence context in system prompts. The study finds that LLMs increase structural complexity (headings, steps, length) under explicit ND instructions, but persona assertion alone fails to suppress harmful behaviors—a critical finding for equitable AI system design.
Researchers at arXiv studied how task phrasing influences the decision-making of large language models, using the iterated prisoner's dilemma as a test case. The findings reveal that LLMs are prone to making presumptions based on how tasks are worded, which can impair their adaptability and reasoning—a safety concern for real-world deployment. Neutral task phrasing significantly reduced these presumptions, suggesting that prompt design is critical for reliable LLM performance.
Researchers propose Comet-H, an AI system that orchestrates language models to generate research software by keeping mathematical theory, code, benchmarks, and documentation synchronized. The framework addresses hallucination and desynchronization failures in LLM-driven development, demonstrating effectiveness through a portfolio of 46 research repositories, with a static-analysis tool reaching F1=0.768 performance.
Research demonstrates that for procedural tasks, simple in-context prompting with complete procedures in the system prompt outperforms complex agent orchestration frameworks like LangGraph and CrewAI. Testing across three domains showed the simpler approach achieved 4.53-5.00 quality scores versus 4.17-4.84 for orchestrated systems, with failure rates 50-76% lower, suggesting advances in frontier LLM capabilities have eliminated the need for external orchestration.
Researchers present LLM+ASP, a framework combining large language models with Answer Set Programming to enable nonmonotonic reasoning without task-specific engineering. The system uses automated self-correction loops where an ASP solver provides structured feedback, demonstrating significant performance improvements over monotonic logic approaches across diverse reasoning benchmarks.
Researchers demonstrate that integrating facial expression analysis into large language model prompts improves empathetic tutoring responses without requiring model retraining. Testing across three major LLM backbones with 960 multi-turn conversations, Action Unit estimation-based conditioning consistently enhanced emotional responsiveness while maintaining pedagogical quality.
Researchers introduce SSAS, a framework that improves LLM consistency for sentiment analysis by applying hierarchical classification and iterative summarization to enforce bounded attention on raw text. Testing on three standard datasets shows the method reduces analytical variance by up to 30%, addressing the fundamental challenge of using non-deterministic LLMs for enterprise-grade analytics.
Researchers introduce DiZiNER, a framework that improves zero-shot named entity recognition by simulating human annotation disagreement processes using multiple LLMs. The approach achieves state-of-the-art results on 14 of 18 benchmarks, closing the performance gap between zero-shot and supervised systems by over 11 percentage points.
Researchers introduce the first benchmark for multicultural text-to-image generation, revealing that state-of-the-art AI models struggle with culturally diverse scenes. The study of 9,000 images across five countries and multiple demographics shows significant performance disparities, with a multi-agent framework using cultural personas demonstrating potential improvements in image quality and cultural accuracy.
Researchers formalize the one-sided conversation problem (1SC), where only one participant's dialogue can be recorded—common in telemedicine, call centers, and smart glasses. The study evaluates methods to reconstruct missing speaker turns and generate summaries from incomplete transcripts, finding that smaller models require finetuning while larger models show promise with prompting techniques.
