Airbnb has developed and deployed Proximity Features, a privacy-compliant system that addresses the cold-start personalization problem by grouping users based on geographic proximity using geo-IP data and adaptive clustering. This approach aggregates signals for groups of around 1,000 users, avoiding the need for persistent individual identifiers and operating within consent-gated privacy controls. Online A/B tests in production show statistically significant increases in bookings, particularly for users lacking recent or any history.
Why it matters: This work provides a scalable, privacy-preserving solution to cold-start personalization with demonstrated real-world impact at a major online platform.
Researchers introduce LiteTopK, a fused GPU kernel designed to accelerate Indexer-TopK operations in long-context sparse attention by leveraging the curse of dimensionality. LiteTopK samples data to estimate score ranges and partitions candidates into bins, reducing memory traffic and overhead while maintaining exact top-k correctness. Experiments show a 1.2x speedup in the prefill stage of GLM 5.2 during real-world deployment, along with lower memory usage.
Why it matters: This work offers a practical advance in the efficiency of sparse attention for large language models, enabling faster and more memory-efficient processing of long contexts.
A new preprint demonstrates that up to 77% of features recovered by sparse autoencoders (SAEs) with high cosine similarity to ground-truth directions are causally inert, meaning the matched atom does not activate when the feature is present. The authors introduce sae-causal-audit, a model-agnostic tool for causal validation, and identify two types of inertness: structural (due to antipodal-pair geometry) and competitive (arising from TopK pathologies in degraded SAEs).
Why it matters: This work challenges the use of correlational metrics alone for evaluating SAE interpretability, showing that high recovery scores can mask a lack of causal relevance, which is important for mechanistic interpretability research.
A new audit of six KV-cache compression methods finds that their performance rankings reverse when compression is performed before seeing the query (query-agnostic) compared to after (query-aware). In the more realistic query-agnostic setting, only KeyDiff consistently outperforms trivial baselines, while the widely used SnapKV method underperforms a simple 'keep start and recent window' baseline.
Why it matters: This study challenges the validity of current evaluation protocols for KV-cache compression, suggesting that some popular methods may be less effective in real-world LLM deployments than previously thought.
A new study demonstrates that applying Principal Component Analysis (PCA) to compress query embeddings can enhance retrieval performance in specialized domains. The approach was tested across 9 dense retrievers and 14 MTEB datasets, showing improved NDCG@10 scores in 75.4% of model-dataset pairs compared to standard embeddings. This method preserves domain-relevant features while removing non-discriminative components, providing a lightweight, training-free alternative to traditional domain adaptation methods.
Why it matters: This technique offers a simple and efficient way to improve dense retrieval in specialized domains without the need for costly annotation or retraining.
A new method equips speculative execution in LLM agents with three online memory systems—a contrastive transition table, episodic memory, and confusion tracker—to improve prediction quality by learning from past agent trajectories. The approach achieves 19–39% relative accuracy improvement on action prediction and up to 2.5× increase on observation prediction tasks, with gains increasing as memory accumulates. All speculation occurs during idle time, resulting in no added wall-clock cost and preserving identical agent trajectories compared to non-speculative execution.
Why it matters: This work demonstrates a way for LLM agents to learn from experience and improve efficiency without sacrificing correctness, potentially advancing the deployment of faster and more capable agentic systems.
Researchers have introduced iTIMO, a dataset designed to address the underexplored task of travel itinerary modification. The dataset is generated by instructing large language models to perturb real-world itineraries using REPLACE, ADD, and DELETE operations, guided by intents such as popularity disruptions, spatial distance, and category diversity. The work also benchmarks state-of-the-art LLMs on this task, revealing their strengths and limitations.
Why it matters: iTIMO provides a comprehensive testbed for developing and evaluating adaptive travel recommendation systems that can handle dynamic and changing user needs.
Researchers introduce SkillSelect-Serve, a framework that recommends reusable agent skills for LLM agents while respecting constraints such as token budget, risk, and tool availability. By profiling skills as structured services and applying a constrained projection method, the system achieves 100% deliverability with minimal loss in hit rate. The approach outperforms traditional Top-k retrieval methods, significantly reducing risk exposure and tool violations.
Why it matters: This work offers a practical advance for deploying LLM agents by enabling reliable, constraint-aware skill selection, which is crucial for efficient operation under real-world resource limits.
Researchers introduce FastAlign, a sparsity-aware framework for optimal transport-based network alignment. FastAlign maintains alignment quality comparable to state-of-the-art methods while significantly improving computational efficiency, achieving up to 9.45x speedup on CPU and 32.54x on GPU. The method leverages mixed sparse-dense operations and custom kernel fusion to address scalability challenges in large-scale network alignment.
Why it matters: This work enables efficient analysis of large-scale networks, addressing a major scalability bottleneck in network alignment tasks relevant to fields such as social network analysis and fraud detection.
A new preprint finds that large language models (LLMs) generate responses that are more concentrated and mainstream than the diverse, long-tail outputs produced by humans. The researchers tested interventions such as increasing temperature sampling, prompting for diverse perspectives, and aggregating outputs from multiple models, finding that these methods can improve diversity, but single-model outputs still fall short of human-level diversity.
Why it matters: The study raises concerns that LLMs could reduce cultural diversity in generated content, which has implications for AI policy and the preservation of democratic values.
Policy & Safety→Official→arXiv Computers and Society
A new benchmark, NOHARM, evaluates 20 large language models (LLMs) and 4 clinical AI tools on 1,100 medical consultation cases, finding that direct use of AI-generated recommendations could result in severe harm in up to 24.6% of cases, with omission errors accounting for over 80% of severe errors. In a randomized study of 101 physicians, AI assistance improved performance, but physicians often omitted valuable AI recommendations, indicating complementary strengths in human-AI teaming.
Why it matters: This study provides the first systematic measurement of clinical safety in LLM-generated medical advice, revealing that widely used AI tools can produce potentially harmful recommendations and highlighting the need for explicit safety evaluation.
Researchers introduce EG-VAR, a Lean 4-based tool-calling architecture that ensures every verified output is grounded in attested tool calls and kernel-checked inference, thereby eliminating unsupported claims. On a subset of TableBench numerical reasoning tasks (n=120), EG-VAR achieves perfect accuracy (120/120) compared to a 95% baseline, and maintains 100% source-faithfulness on stress tests where baselines drop to 80-90%. The system also provides explicit audit trails for abstentions and formalization errors.
Why it matters: EG-VAR offers a practical and auditable approach to eliminating LLM hallucination in empirical inference, potentially transforming trustworthiness in high-stakes AI applications.
MESH is a unified retrieval scaling framework designed to address the scaling bias of heterogeneity in large-scale recommendation systems. Through a modularized architecture with gated bias correction, MESH partitions the feature space to reduce interference between different content types, leading to a 14x improvement in the power-law scaling exponent for fresh items. In online evaluations on Pinterest's Related Pins platform, MESH demonstrated a +5.5% lift in fresh-item repins, a 55% improvement in funnel efficiency, and a +0.46% improvement in user retention.
Why it matters: MESH represents a significant advance in unifying fragmented retrieval systems, improving scalability and performance for diverse content in large-scale recommendation platforms.
A new preprint compares large language models (LLMs) such as GPT, Twitter-roBERTa, and LLaMA to traditional machine learning methods for analyzing open-ended survey responses. The study finds that LLMs achieve higher classification accuracy, especially in sentiment and thematic analysis, but exhibit significant variation in consistency and the explicitness of their reasoning. These results highlight important trade-offs between predictive performance and interpretability in large-scale qualitative research.
Why it matters: The study offers practical insights for researchers seeking to balance automation with interpretive rigor when applying LLMs to qualitative data analysis.
AgentSociety 2 is a new research environment that integrates large language model (LLM) agents as both AI social scientists and simulated participants, automating the end-to-end workflow of social science research. The system enables hypothesis generation, experiment design, simulation execution, result interpretation, and manuscript drafting across micro, meso, and macro social scenarios. It demonstrates the ability to reproduce qualitative patterns from prior studies and supports large-scale, auditable simulations.
Why it matters: This work advances computational social science by providing a scalable, reproducible, and auditable platform for automating complex social experiments with AI agents.
A new preprint examines the use of large language models (LLMs) for content moderation through 'policy-as-prompt' methods, where moderation policies are given to LLMs as natural-language prompts. The authors argue that this approach introduces specific risks and harms, and that simply writing prompts is not sufficient for effective or meaningful community governance. They propose several considerations for improving prompt governance but conclude that prompt-writing alone cannot ensure robust moderation outcomes.
Why it matters: This research highlights important limitations and governance challenges for AI-driven, prompt-based content moderation systems as their use expands in online communities.
A new preprint compares institutional and course-level generative AI policies at U.S. research-intensive universities. The study finds that while institutions tend to be more supportive of GenAI use, course-level guidance in computing education remains cautious. The authors propose an instructor-centered framework to guide future GenAI adoption in courses.
Why it matters: This research highlights a disconnect between university-wide AI policies and classroom practices, offering a framework to help computing educators navigate GenAI adoption.
AutoTrace is an agentic pipeline that uses LLM agents to explore code property graphs and localize vulnerability triggers, even when they are located outside patched functions. On the InterPVD benchmark, AutoTrace achieves 75.0% VulnHit and 80.8% FuncHit, surpassing previous methods. The authors also introduce SinkTrace-Bench, a dataset of 1,542 source-to-sink causal chains, which reveals that current frontier LLMs struggle with causal reasoning in vulnerability analysis.
Why it matters: This work advances automated vulnerability analysis by enabling interprocedural trigger localization and provides a new benchmark that highlights the causal reasoning limitations of current LLMs.
A new preprint introduces StableAML, a machine learning framework for detecting money laundering in stablecoin transactions on Ethereum. The study finds that domain-informed tree ensemble models outperform graph neural networks in identifying suspicious wallets, and can distinguish between behavioral patterns of cybercrime syndicates and sanctioned entities. The approach is designed to support compliance with emerging regulations such as the EU's MiCA and the U.S. GENIUS Act.
Why it matters: This work presents a novel, interpretable method for high-precision behavioral classification in stablecoin anti-money laundering, potentially improving compliance and reducing unjustified asset freezes under new regulatory frameworks.
A new preprint introduces the agent:// URI scheme, designed to decouple agent identity from network location in multi-agent systems. The scheme incorporates trust roots, hierarchical capability paths, and sortable unique identifiers, with cryptographic attestation using PASETO tokens. Evaluation demonstrates full capability expressiveness on 369 tools, perfect discovery precision across 10,000 agents, and sub-5-microsecond performance, suggesting a robust and scalable approach to agent identity and discovery.
Why it matters: This work proposes a novel, decentralized solution to agent identity and discovery, potentially enabling more robust and scalable multi-agent ecosystems.