TechniqueRLHF / Alignment8 recent entries12 Oct 2026Neural networks are more modular than single neurons suggestarXiv:2610.12842v1 Announce Type: new Abstract: Neural networks routinely develop representations in which individual neurons contribute to multiple distinct functions, leading to the widespread view →12 Oct 2026Finite-Sample Guarantees for Fully Discrete Conditional Diffusion ModelsarXiv:2610.13851v1 Announce Type: cross Abstract: We establish finite-sample guarantees for conditional diffusion models learned from data and implemented with finitely many denoising steps. Our appro
TechniqueRAG8 recent entries9 Oct 2026Is Memorization Context-Sensitive? Prefix-Based Extraction Beyond Isolated PrefixesarXiv:2610.12085v1 Announce Type: new Abstract: Large language models (LLMs) can expose memorized training sequences under prefix-based extraction: given a prefix from a training example, the model ma→9 Oct 2026Beyond Sequences: Distilling Structured Decision Memory for LLM RecommendationarXiv:2610.11501v1 Announce Type: new Abstract: Despite the adoption of large language models (LLMs) in recommendation systems, prevailing approaches mostly model single-type behaviors (e.g., views or→12 Oct 2026Towards Enriching Design-Time Models with Evidence from Simulation and Operational ActivitiesarXiv:2610.13456v1 Announce Type: new Abstract: Cyber-Physical Systems are developed through multiple model based activities, comprising several stages of development. As the development progresses th→12 Oct 2026PRIME: Pre-Retrieval Index Modeling Enhancement for Structured Document Retrieval-RealarXiv:2610.13296v1 Announce Type: cross Abstract: High-precision retrieval over external knowledge bases is essential for LLM agents to operate effectively. However, existing Information Retrieval (IR→12 Oct 2026Model-agnostic Knowledge Augmented Machine Learning for Weakly Annotated Bioimaging DataarXiv:2610.13637v1 Announce Type: cross Abstract: Machine learning models are increasingly used for automated feature extraction from weakly annotated bioimaging data, such as microscopy and cytomorph→12 Oct 2026MHR-Router: Multimodal Heterogeneous Retrieval-Enhanced Adaptive Routing for VLM InferencearXiv:2610.13278v1 Announce Type: new Abstract: Vision-language models (VLMs) exhibit complementary capabilities, yet selecting the most suitable expert for a given image-question query remains challe→12 Oct 2026KV-Repurpose: A Testbed for Evaluating Non-Prefix KV-Cache ReusearXiv:2610.13060v1 Announce Type: new Abstract: KV-cache reuse in large language model serving is commonly limited to exact prompt prefixes, preventing reuse when the same retrieved document appears a→12 Oct 2026CRM: Retrieval Model with Controllable ConditionarXiv:2412.13844v2 Announce Type: replace-cross Abstract: Retrieval models for recommendation need to account for outcomes such as watch time while searching large item corpora efficiently. We present
TechniqueAgents8 recent entries9 Oct 2026RAG-Stress: Probing the Limits of Evidence Reliance in Retrieval-Augmented GenerationarXiv:2610.11183v1 Announce Type: cross Abstract: Following retrieved evidence does not guarantee factual correctness: misleading evidence can induce a model to replace an answer it previously gave co→9 Oct 2026No Distillation Needed: Single-Pass Real-Time Talking Heads via Acausal Noise ShapingarXiv:2610.11070v1 Announce Type: new Abstract: Audio-driven facial animation underpins real-time avatars, telepresence, and embodied virtual agents. And it must run online: each frame emitted from au→9 Oct 2026Memory Type Varies: Empowering LLM Agents for Long-Term Memory with Diverse StrategiesarXiv:2610.11573v1 Announce Type: new Abstract: The memory capabilities of Large Language Models (LLMs) have garnered increasing attention recently. Despite great success achieved, existing retrieval-→9 Oct 2026Caught in the Act: Probes Effectively Detect Sabotage and Catch Unverbalized DeceptionarXiv:2610.12445v1 Announce Type: cross Abstract: Recent incidents have highlighted the challenge of monitoring LLM agents and the danger of models deceiving people. We show that white-box deception d→9 Oct 2026Ai2 at COLM 2026: Open research, from models to agents for scienceAi2 joined COLM 2026 to share advances in open language models, training infrastructure, and AI for science, including Olmo Hybrid, Asta, and Bolmo, while connecting with the broader research communit→10 Oct 2026Pay close attention to custom harnesses. This is a super interesting paper showing the impact of training inside the harness. Qwen3-14B goes…Pay close attention to custom harnesses. This is a super interesting paper showing the impact of training inside the harness. Qwen3-14B goes from 22.2% to 54.8% on Spider 2.0-SQLite when it is trained→12 Oct 2026Returning to Development States: When Full Sandbox Snapshots Help Coding AgentsarXiv:2610.13841v1 Announce Type: new Abstract: Coding agents often make useful progress before they fail, but retry-from-scratch discards the machine state created along the way. This makes failure r→12 Oct 2026PRIME: Pre-Retrieval Index Modeling Enhancement for Structured Document Retrieval-RealarXiv:2610.13296v1 Announce Type: cross Abstract: High-precision retrieval over external knowledge bases is essential for LLM agents to operate effectively. However, existing Information Retrieval (IR
TechniqueFine-tuning8 recent entries12 Oct 2026Robust Langevin Sampling under State-Dependent Heavy-Tailed Noise with Applications to LLM Fine-TuningarXiv:2610.12865v1 Announce Type: new Abstract: Existing convergence guarantees for non-log-concave sampling with stochastic gradient Langevin Monte Carlo (SG-LMC) typically assume globally Lipschitz →12 Oct 2026Not All Dimensions Are Equal: Outlier-Aware Attention Linearization with Grouped Quadratic KernelsarXiv:2610.13674v1 Announce Type: cross Abstract: Linearizing softmax attention offers a route to efficient Transformer inference, but preserving pretrained attention remains challenging. We identify →12 Oct 2026Monarch-MoE: Structured Experts for Multimodal Continual Instruction TuningarXiv:2610.13215v1 Announce Type: new Abstract: Learning new tasks in multimodal continual instruction tuning can cause catastrophic forgetting of previously acquired knowledge. Although multi-expert →12 Oct 2026Mesh-Native Tokenisation: Retrofitting Foundation Models to Unstructured DataarXiv:2610.12618v1 Announce Type: cross Abstract: Physics foundation models are an increasingly popular approach for building versatile emulators for spatiotemporal systems. They require tokenisation →12 Oct 2026Golden Anchor for Video GenerationarXiv:2610.13217v1 Announce Type: cross Abstract: Video generation supports digital content creation, with applications in virtual avatars, AI-driven filmmaking, interactive game engines, and world mo→12 Oct 2026Do Prompt Optimizers Break Requirements They Are Not Scored On? A Preregistered StudyarXiv:2610.13287v1 Announce Type: new Abstract: Prompt optimizers rewrite a prompt to raise one target metric, but prompts also state requirements the metric does not score, such as an output format, →12 Oct 2026AI Models Can Provably Hide Arbitrary CapabilitiesarXiv:2610.13928v1 Announce Type: new Abstract: Capability evaluations aim to surface dangerous model behaviors before deployment, but their reliability depends on the assumption that hidden capabilit→12 Oct 2026A Critical Look at Power Sampling when Test Time Sharpening LLMsarXiv:2610.13653v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has substantially improved LLM reasoning, yet growing evidence suggests it mainly performs distr
TechniqueMultimodal8 recent entries12 Oct 2026Neural networks are more modular than single neurons suggestarXiv:2610.12842v1 Announce Type: new Abstract: Neural networks routinely develop representations in which individual neurons contribute to multiple distinct functions, leading to the widespread view →12 Oct 2026Multimodal Image-based Robotic Assistance and Guidance Environment (MIRAGE): A Novel Unified Framework for Surgical Navigation and VisualizationarXiv:2610.12780v1 Announce Type: new Abstract: This paper presents the computational architecture and quantitative preclinical evaluation of MIRAGE, a multimodal image-guided robotic surgery framewor→12 Oct 2026Monarch-MoE: Structured Experts for Multimodal Continual Instruction TuningarXiv:2610.13215v1 Announce Type: new Abstract: Learning new tasks in multimodal continual instruction tuning can cause catastrophic forgetting of previously acquired knowledge. Although multi-expert →12 Oct 2026Model-free Bayesian Clustering and Multimodal Posterior SummarisationarXiv:2610.13397v1 Announce Type: cross Abstract: We develop a model-free, noise-aware Bayesian clustering framework that constructs a martingale posterior over clusterings from a predictive Bayesian →12 Oct 2026MHR-Router: Multimodal Heterogeneous Retrieval-Enhanced Adaptive Routing for VLM InferencearXiv:2610.13278v1 Announce Type: new Abstract: Vision-language models (VLMs) exhibit complementary capabilities, yet selecting the most suitable expert for a given image-question query remains challe→12 Oct 2026Kairos: Toward a Byte-Level Foundation Model for Every ModalityarXiv:2610.12495v1 Announce Type: cross Abstract: Small language models trained on a single local machine leave little room for architectural trial and error. Kairos combines bidirectional gated-Delta→12 Oct 2026GlucoPRISM: Protocol-Supervised Pretraining for Cross-Cohort Glucose TransferarXiv:2610.12528v1 Announce Type: new Abstract: Continuous glucose monitoring (CGM) foundation models encode each day as one vector that mixes who the person is, what happened that day and how the sen→12 Oct 2026Energy-Efficient Spiking Neural Networks for Sensing-Aided Beam PredictionarXiv:2610.12507v1 Announce Type: cross Abstract: Sensing-aided beam prediction can reduce the beam-training overhead of millimeter-wave (mmWave) systems by leveraging environmental information from c
TechniqueSafety8 recent entries9 Jun 2026What the Eyes See, the LLMs Miss: Exploiting Human Perception for Adversarial Text AttacksarXiv:2606.09700v1 Announce Type: cross Abstract: Large language model (LLM)-powered content moderation systems have become a critical defense against harmful online content. However, these systems pr→24 Jun 2026The best way to understand a complex system is via edge cases and failure modes, because they define the contour of the system.Edge cases and failure modes reveal the fundamental boundaries and constraints of complex systems more effectively than typical operations, making them valuable for understanding system behavior and l→30 Jun 2026Defending Against Harmful Supervision Hidden in Benign SamplesarXiv:2606.30263v1 Announce Type: cross Abstract: Existing defenses are effective when harmful content is explicitly mixed into downstream fine-tuning data, but crafted samples can instead hide harmfu→4 Aug 2026Training-Free Out-of-Distribution Detection for Pathology Whole-Slide ImagesarXiv:2608.01407v1 Announce Type: new Abstract: Safe deployment of AI methods in medicine requires robust guardrails that detect when input data deviate from the training distribution to ensure that m→10 Sept 2026You Are What You Read: Misalignment via In-Context Persona InductionarXiv:2609.06851v1 Announce Type: cross Abstract: Broad misalignment has been produced by finetuning on narrow data, harmful or benign, and in context only by demonstrations of the undesirable behavio→12 Sept 2026Architecting the Secure AI-SOC: A Neurosymbolic Framework for Pipeline Integrity and Threat MitigationarXiv:2609.10707v1 Announce Type: cross Abstract: The integration of Large Language Models (LLMs) into Security Operations Centers (SOCs) streamlines threat intelligence but introduces critical vulner→29 Sept 2026Quantization Error Is Spectrally Flat: A Single Random Probe Is a Calibrated, Data-Free Sensitivity Estimator, with Application to Budget-Targeted Mixed-Precision QuantizationarXiv:2609.33923v1 Announce Type: cross Abstract: A single random Gaussian probe gives an unbiased estimate of the squared Frobenius norm of a layer's quantization error. The estimator is well-behaved→1 Oct 2026Examining Variation in How Guided AI Tutors Resolve Student ImpassesarXiv:2609.38346v1 Announce Type: new Abstract: When a student is stuck, a tutor faces the assistance dilemma: help given too early can hinder productive struggle, while help withheld too long leaves