Latest Prompting & ICL Research Papers

The newest Prompting & ICL papers from across the field — arXiv, NeurIPS, CVPR, Nature, and more — refreshed daily and ranked by relevance. Distill AI tracks Prompting & ICL so you don’t have to: get the standout work delivered to your inbox every morning, with 2-sentence summaries and the option to chat with any paper.

Recent papers

• Social Capital and Regeneration in Historical Areas: Case Study of Jianguo Men Market, Xi’an, China
  Min Wang · University of Liverpool · Jan 1, 2029

  In recent years, China has increasingly adopted more inclusive and participatory approaches to urban regeneration, prompted by constrained government funding and the limitations of traditional top-down models. Although existing literature l…

• Moral Consistency Variance: A Pilot Benchmark for Decision Stability under Moral Prompt Perturbations in Large Language Models
  Qiao Liang · OpenAlex · Dec 31, 2026

  Static moral question-answering benchmarks do not test whether a model's decision distribution remains stable when the same dilemma is rephrased without changing the underlying facts. This paper introduces Moral Consistency Variance (MCV), …

• RAG-Safety-Bench: Reliable Evaluation of Retrieval-Augmented LLM Safety
  Adithiyan Rajan Indira Saravanan, Kathleen C. Fraser · arXiv · Sep 10, 2026

  Allowing large language models (LLMs) to retrieve information from a set of trusted documents can increase reliability and reduce hallucination. However, recent work has demonstrated that retrieval-augmented generation (RAG) can have uninte…

• SWRouter: Similarity-Contractive Window Routing for Multi-Turn Large Language Model Conversations
  Yu Wang, Yuchen Li, Rui Kong, Xinran Chen et al. · arXiv · Sep 10, 2026

  Large language models exhibit complementary strengths, motivating routing methods that dispatch each query to the most suitable model. Although existing routers are effective in single-turn settings, they do not directly transfer to multi-t…

• Rosetta at AlexandriaX-2026: LoRA-Adapted NileChat for Context-Aware Dialectal Arabic Dialogue Translation
  Nada Esmaeil, Fathima Rena, Sibi Subhash, Osama Elgendy et al. · arXiv · Sep 9, 2026

  This paper describes the Rosetta system for Subtask 1 (Context-Aware English-to-Dialectal Arabic Dialogue Translation) of the AlexandriaX shared task, participating in both constrained and unconstrained tracks. The approach fine-tunes a LoR…

• Why Is Video Still So Expensive? A Survey of Inference-Efficiency Mechanisms in Video and Audiovisual LLMs
  Killian Steunou, Yannis Tevissen, Mounîm A. El Yacoubi · arXiv · Sep 9, 2026

  Video understanding has rapidly evolved toward video large language models (VideoLLMs): systems that couple video representations with pretrained large language models and condition generation on a textual prompt. Their strong performance o…

• KVShareArena: KV-Cache Reuse Across Contexts and Model Checkpoints
  Xi Shi, Qian Lou · arXiv · Sep 9, 2026

  LLM serving systems already reuse KV caches, but only when the reused text sits at the very start of the prompt. Two growing workloads break this condition: a retrieval-augmented generation server assembles a different set of retrieved chun…

• Two-Token Features and Small-Large Ensembles for VLM Hallucination Detection
  Eli Schwartz · arXiv · Sep 9, 2026

  We present our system for the SHROOM-Visions 2026 shared task on character-level VLM hallucination detection. A small ($4$B-parameter) VLM is fine-tuned as a per-token classifier reading a two-token feature from its own hidden states, and i…

• Influence Score and Transformers interpretability: Measure of the Effective Impact of Attention Heads at inference time
  Lisa Bouger, Yannick Teglia, Philippe Loubet Moundi · arXiv · Sep 4, 2026

  We propose an influence score to quantify the contribution of attention heads to classification decisions in Transformer-based models designed for prompt injection detection. The score combines directional influence on the logits with struc…

• ESPO: Error-Structured Prompt Optimization via Diagnose, Diversify, and Stabilize
  Lihao Liu, Peng Tang, Kunwar Yashraj Singh, Shabnam Ghadar · arXiv · Sep 3, 2026

  Evolutionary prompt optimizers such as GEPA suffer from prompt bloat: each iteration appends rules and caveats, producing prompts up to 3$\times$ longer yet no more accurate. We trace this to three deficiencies - incomplete error observatio…

• IchthyoNoma: Nomenclature and Context Sensitivity of Zero-Shot Biological Vision--Language Models for Bangladeshi Freshwater Fish Recognition
  Nazim-E-Alam, Tarek Rahman, Md Kishor Morol · arXiv · Sep 3, 2026

  Zero-shot vision-language models (VLMs) are increasingly used as training-free species recognizers, but reported accuracy can reflect more than visual species knowledge. We audit CLIP, BioCLIP, BioCLIP2, and a multilingual Jina CLIP v2 cont…

• DKL: Decoupled Knowledge Learning for Instruction-Tuned Language Models
  Kushagra Bhushan, Meghanadh Pulivarthi, Sai Krishna Reddy Sathi, Gaurav Pandey et al. · arXiv · Sep 2, 2026

  RAG has become the de facto method for incorporating new, corpus-specific knowledge into an instruction following LLM (Instruct LLM). Although RAG-based prompting improves factual grounding, it fails when retrieval is incorrect or incomplet…

• From Confusion to Clarity: Confusion-Aware Retrieval and Knowledge Injection for Text Classification
  Manish Gupta, Chaitanya Giri, Jayasimha Talur · arXiv · Sep 1, 2026

  Large language models (LLMs) struggle to classify text into taxonomies with many semantically similar labels, as the distinctions are domain-specific and not captured by pre-training. To handle large label spaces, a common approach retrieve…

• SDARE-Bench: Evaluating Large Language Models on Conversational Stigma Detection and Response in Dyadic and Group Dialogue
  Stephanie Fong, Yiwen Jiang, Zimu Wang, Hongxi Yang et al. · arXiv · Sep 1, 2026

  Large Language Models (LLMs) are increasingly used in advice seeking and decision making that may affect social judgements. Despite stigma's profound effects on people and communities, benchmarks remain scarce. Existing general-domain evalu…

• Are These Modules Worth Their Cost? A Paradigm-Level Accuracy-Cost Analysis of In-context Learning Text-to-SQL
  Jiayan Lin, Yujia Liu, Zijin Hong, Zheng Yuan et al. · arXiv · Aug 28, 2026

  Recent advances in in-context learning (ICL) text-to-SQL have substantially improved execution accuracy on public benchmarks by assembling increasingly elaborate pipelines around the base generator, yet existing studies typically report agg…

• The Invisible Editorial Layer: Formalizing Undisclosed Inference-Time Steering, Probability Placement, and the Attribution Problem in Deployed Language Models
  Augusto Camargo · arXiv · Aug 25, 2026

  Large language models (LLMs) are commonly evaluated under the assumption that their observable behavior is primarily determined by model weights, training data, alignment procedures, and user prompts. This view is incomplete. Modern inferen…

• Prompt-Model Interaction Reaches the Fixed Points: A deterministic, task-free structural readout -- and the factorizations of it that failed
  Nicolás Vera Zúñiga · arXiv · Aug 21, 2026

  That a prompt's effect is not a property of the prompt is established: prompts optimised for one model degrade on another, and rankings reorder under neutral reformatting. That evidence is about task accuracy, which cannot say whether the i…

• No PUN Intended: Plausible Unknown Names for Person-Centred LLM Evaluation
  Dimitri Staufer, David Hartmann, Ibrahim Baroud · arXiv · Aug 21, 2026

  Person names are widely used as prompt variables in LLM evaluations of factuality, privacy leakage, bias and abstention, but when a name's evidential status is uncontrolled, measurements may conflate memorisation, retrieval, name priors and…

• PromptResponse: Optimizing Prompts for LLM Coding Tasks
  Erik Thureck, Robert Kühnen, Tim Jacobowitz · arXiv · Aug 21, 2026

  Large language models (LLMs) are increasingly used in research workflows and software development pipelines, yet their output remains sensitive to input prompt variations. This paper presents $\unicode{x00AB}$PromptResponse$\unicode{x00BB}$…

• Training Chemical Plausibility-Aware Large Language Models for Single-Step Retrosynthesis
  Bogdan Zagribelnyy, Ivan Ilin, Nikita Bondarev, Maksim Kuznetsov et al. · arXiv · Aug 19, 2026

  Single-step retrosynthesis is a central component of computer-aided synthesis planning, yet its intrinsically one-to-many nature is poorly captured by single-answer evaluation and benchmarking protocols. To address this, we introduce Top-K …

• On the Fragility of Self-Improving Agents: Variance, Task Order, and Underspecification
  Qinyuan Ye, Yu Li, Yada Pruksachatkun, Jiaxin Zhang et al. · arXiv · Aug 18, 2026

  Memory-based self-improving agents--those that learn from an online stream of tasks and improve over time by maintaining a textual memory bank--have shown great promise in recent literature. However, the reliability aspects of these methods…

• BayesPrompt: human readable prompts that make sense
  Franky Kevin Nando Tezoh, Ali Hussaini Umar, Alessandro Laio, Guido Sanguinetti et al. · arXiv · Aug 18, 2026

  Reconstructing prompts that can elicit a desired answer or behaviour in an LLM is an open and important research topic. Optimisation methods which aim at minimising the perplexity of a given answer, however, consistently yield so-called pse…

• An Empirical Study of Reward Specification and Benchmark Reliability in GRPO-based LLM Unlearning
  Rubén Balbastre, Juan Manuel Orduña, Mariano Pérez · arXiv · Aug 18, 2026

  Practical LLM unlearning is usually evaluated through two objectives: suppress target-specific knowledge and preserve non-target utility. In generative QA, this leaves a third behavior underspecified: when a target-adjacent prompt admits a …

• Policy Iteration with Human Feedback: Bringing Post-Training RL to In-context Learning
  Minh-Ha Nguyen, Cathy Shyr · arXiv · Aug 17, 2026

  Generative pretraining established reusable task representations; later work on language-based task conditioning and in-context learning showed that a fixed model could adapt its behavior from instructions and demonstrations. Policy Iterati…

• When Do Explanations Help In-Context Learning? A Comparative Study of Natural Language Explanation Types and Faithfulness
  Mahdi Dhaini, Adam Dejl, Juraj Vladika, Volkan Özer et al. · arXiv · Aug 17, 2026

  Natural language explanations (NLEs) are increasingly used as inputs, for example, as few-shot rationales that influence model behavior in in-context learning (ICL). However, it remains unclear how different types of NLEs compare in their e…

• DSPrompt: Dynamic Soft Prompt Defense Against M-RAG Corruption
  Chang Liu, Yuni Lai, Mingyue Cui, Cong Tian et al. · arXiv · Aug 17, 2026

  Multimodal Retrieval Augmented Generation (M-RAG) is increasingly vulnerable to adversarial attacks where malicious data are crafted to produce embeddings that align with benign entries in the vector space, deceiving retrieval and inducing …

• Split the Labor: Separating Evidence Interpretation from Decision Aggregation
  Zhelun Wu · arXiv · Aug 14, 2026

  Systems that ask a language model to reach a conclusion from many sources usually concatenate them into one prompt. This conflates two operations with different requirements. Interpreting a source rewards capacity and context. Combining int…

• Envs-FORGE: Frontier-Optimized Reward-Grounded Environment Synthesis for Agent RL
  Xiaojun Wu, Cehao Yang, Honghao Liu, Xueyuan Lin et al. · arXiv · Aug 14, 2026

  Reinforcement learning (RL) for terminal agents needs executable training environments with reliable rewards and useful difficulty. Fixed recipes such as few-shot, Self-Instruct, and Evol-Instruct apply the same prompting policy to every se…

• Seeing Red, Thinking Bad: Color Bias in Vision Language Models
  Kohsuke Ide, Ryousuke Yamada, Yoshihiro Fukuhara, Hirokatsu Kataoka et al. · ICPR 2026 · Aug 14, 2026

  Vision language models (VLMs) are increasingly used in industrial decision-making systems, such as recruitment support and recommendation. This motivates careful analysis of how VLMs process visual and textual information. In this work, we …

• MINT: A Universal Zero-Shot Predictor for Transaction Data
  Parameswaran Kamalaruban, Viktor Drobnyi, Maeve Madigan, Julia Rozanova et al. · arXiv · Aug 14, 2026

  Banks analyse sequential financial transaction data to perform many tasks, including fraud prevention, credit risk assessment and offer personalization. To improve the predictive accuracy of these tasks, Payments Foundation Models encode tr…