PreprintNot peer reviewedArtificial IntelligenceStudy finds LLM safety refusals can be bypassed by suppressing silent featuresOriginal title: Don't Judge an LLM Only by Its Activations: Discovering Suppressed Safety Features via Counterfactual Activation PotentialA preprint reports that jailbreaks can switch off hidden safety-critical features in language models rather than only activating harmful ones.Swadesh Swain et al. · Added 18 Oct 2026
PreprintNot peer reviewedFrom Google DeepMindArtificial IntelligenceA preprint reports online alignment methods beat offline ones in RLHF experimentsOriginal title: Understanding the performance gap between online and offline alignment algorithmsA preprint finds on-policy sampling, not data coverage or quality, likely explains why online RLHF methods outperform offline ones.Added 18 Oct 2026 · Released May 2024
PreprintNot peer reviewedFrom BaiduArtificial IntelligencePreprint proposes constraint-level checks for LLM-generated vehicle routing codeOriginal title: Beyond Objective Equivalence: Constraint Injection for LLM-Based Optimization Modeling on Vehicle Routing ProblemsA preprint introduces constraint injection, a way to check LLM-generated routing code for missing or extra constraints, not just correct objective values.Added 18 Oct 2026
PreprintNot peer reviewedFrom Almawave LabsArtificial IntelligenceA preprint reports PII extraction attacks transfer across languages in LLMsOriginal title: Cross-Lingual Transferability of Training Data Extraction Attacks to Recover Memorized PIIA preprint finds that prompts in other languages can still pull out English-memorised personal data from large language models.Added 18 Oct 2026
PreprintNot peer reviewedArtificial IntelligencePreprint describes sleeper-style attacks that hide in AI agent memory and skillsOriginal title: Plant, Persist, Trigger: Sleeper Attack on Large Language Model AgentsA preprint reports that adversarial content can lie dormant in an LLM agent's state and later trigger harm via an unrelated, benign request.Yongxiang Li et al. · Added 18 Oct 2026
PreprintNot peer reviewedFrom Relty TechnologyArtificial IntelligencePreprint proposes token-selection method to linearise LLM attention more efficientlyOriginal title: STILL: Selecting Tokens for Intra-Layer Hybrid Attention to Linearize LLMsA preprint describes STILL, a method that picks important tokens for full attention and summarises the rest linearly, aiming to cut compute costs.Added 18 Oct 2026
PreprintNot peer reviewedFrom SnowflakeArtificial IntelligenceA preprint describes EviLLM, an attack that makes AI coding tools quietly inject vulnerabilitiesOriginal title: Beware EviLLM: Enabling Vulnerability Injection via Large Language ModelsA preprint reports a third-party attack, EviLLM, that manipulates ChatGPT-based code generation to inject vulnerabilities most study participants failed to notice.Added 18 Oct 2026
PreprintNot peer reviewedFrom MicrosoftArtificial IntelligencePreprint finds sliding-window attention often beats retrofitted linear attention in LLMsOriginal title: Sliding-window beats linear attentionA preprint reports that simple sliding-window attention with sinks matches or outperforms retrofitted linear attention, especially on long-context tasks.Added 18 Oct 2026
PreprintNot peer reviewedFrom MartianArtificial IntelligenceSteering vectors can cut self-preference bias in LLM judges, preprint reportsOriginal title: Breaking the Mirror: Activation-Based Mitigation of Self-Preference in LLM EvaluatorsA preprint reports that activation steering vectors reduce unjustified self-preference in LLM evaluators far more than prompting or DPO.Added 18 Oct 2026
PreprintNot peer reviewedFrom Microsoft ResearchArtificial IntelligenceResearchers propose training safety routers on data isolated from harmful contentOriginal title: Dialectics of Alignment: Harnessing Unsafe Knowledge for Dynamic Safety RoutingA preprint describes SafeMoE, a system that isolates harmful knowledge into expert modules and routes around it to give safer, more informative answers.Added 18 Oct 2026
PreprintNot peer reviewedArtificial IntelligenceA preprint reports a training method for more calibrated LLM-based numeric predictionsOriginal title: Reinforcement Learning over Predictive Distributions for LLM RegressionA preprint introduces a reinforcement learning reward that trains LLMs to produce better-calibrated predictive distributions alongside more accurate point estimates.Jungsoo Park et al. · Added 18 Oct 2026
PreprintNot peer reviewedFrom NTTArtificial IntelligenceResearchers report prompt injection can fool LLM-based phishing detectorsOriginal title: Clouding the Mirror: Stealthy Prompt Injection Attacks Targeting LLM-based Phishing DetectionA preprint finds that hidden instructions in phishing sites can mislead multimodal LLM phishing detectors, and tests a defence to reduce this.Added 18 Oct 2026
PreprintNot peer reviewedArtificial IntelligenceResearchers report gains from delaying KV cache eviction until first decode stepOriginal title: DeferKV: Rethinking Eviction Timing for One-Shot KV Cache CompressionA preprint proposes waiting until the first decoding step before trimming the KV cache, reporting consistent gains over prompt-only compression methods.Zhe Wang et al. · Added 18 Oct 2026
PreprintNot peer reviewedArtificial IntelligenceA preprint proposes a theory for when external guidance helps LLM reasoning trainingOriginal title: When Does External Guidance Help LLM Reasoning? A Bias-Variance Theory of Guidance-Augmented GRPOA preprint offers a bias-variance theory and optimal weighting rule for guidance-augmented reinforcement learning on reasoning tasks.Sofia Torres et al. · Added 18 Oct 2026 · Released Jul 2026
PreprintNot peer reviewedArtificial IntelligenceA preprint describes a cost-aware agent system for proving code correctness in Lean 4Original title: Cost-Efficient Theorem Proving via Agent Orchestration in Program VerificationResearchers report a theorem-proving agent that balances solve rate against token cost across Lean 4 verification benchmarks.Shuangjie Yao et al. · Added 18 Oct 2026
PreprintNot peer reviewedArtificial IntelligenceA preprint benchmark tests whether LLMs judge microcontroller electrical safety correctlyOriginal title: ElecSafety: Diagnosing Large Language Model Safety Judgments for Microcontroller BoardsA preprint introduces ElecSafety, a benchmark finding that LLMs often give correct safety labels for the wrong reasons.Linjian Yang et al. · Added 18 Oct 2026
PreprintNot peer reviewedArtificial IntelligenceStudy questions how hallucination benchmarks label LLM answers as wrongOriginal title: The Labeling Problem in Hallucination Detection Benchmarks: An Empirical EvaluationResearchers find that automated labelling methods for hallucination detection often disagree with human judgement and with each other.Jorma Valjakka et al. · Added 18 Oct 2026
PreprintNot peer reviewedFrom IBM ResearchQuantum ComputingA preprint reports a shallow, verifiable quantum sampling test under lattice assumptionsOriginal title: Verifiable quantum advantage in extremely low depthA preprint describes a sampling task solvable by very shallow quantum circuits yet verifiable classically and hard under lattice assumptions.Added 18 Oct 2026
PreprintNot peer reviewedArtificial IntelligencePreprint describes tile-aligned mixed-precision quantization for faster LLM inferenceOriginal title: AlignQuant: Tile-Aligned Mixed-Precision Quantization for Efficient LLM GenerationA preprint reports a quantization method that aligns precision choices with GPU tiles to speed up LLM generation while preserving quality.Hanzhi Zhang et al. · Added 8 Oct 2026
PreprintNot peer reviewedQuantum ComputingA preprint audits nine PQC libraries and builds a hybrid-by-default Python toolkitOriginal title: quantum-safe: Bridging the Post-Quantum Production Gap with a Hybrid-by-Default Python Cryptography LibraryA preprint reports an external-anchored audit of post-quantum crypto libraries and introduces a hybrid Python library to close production gaps.Animesh Shaw · Added 8 Oct 2026
