← Home

Blog

AI Research, Security & Technology

Building Culturally Specific Stereotype Datasets with LLMs

12 July 2026

A new human-LLM framework constructs EspanStereo, the first native Spanish stereotype dataset, revealing significant cross-country bias variation in LLMs.

Bias & FairnessMultilingual NLPDataset ConstructionLLM Evaluation

Mesh LLM: Distributed Inference Over a P2P QUIC Mesh

12 July 2026

A technical analysis of Mesh LLM, which pools heterogeneous GPUs across iroh endpoints to serve large model inference without centralised infrastructure.

Distributed SystemsLLM InferencePeer-to-PeerQUICAI Infrastructure

From Solvers to Research Agents: LLMs in Formal Mathematics

11 July 2026

A position paper argues that AI mathematics systems must shift from competition solvers to research agents, reviewing formal proof methods and frontier results.

AI ResearchMathematicsLLMsFormal VerificationTheorem Proving

ReCoLoRA: Spectral Consolidation for Continual LLM Fine-Tuning

11 July 2026

ReCoLoRA addresses catastrophic forgetting in sequential LLM fine-tuning using spectral adapter initialisation and recursive weight consolidation across tasks.

Continual LearningPEFTLoRALLM Fine-TuningCatastrophic Forgetting

Self-Distillation for Web Search Agents Without Teacher Models

11 July 2026

DeepSearch-World offers a verifiable environment and self-distillation framework enabling web agents to improve from their own experience, reaching 61.5% on GAI

LLM AgentsReinforcement LearningWeb SearchSelf-ImprovementNLP

Agentic Harnesses for ARC-AGI: 67% at Under $1 Per Task

10 July 2026

A new paper shows agentic decomposition on ARC-AGI-1 reaches 67.25% pass@2 at $0.62 per task using no benchmark-specific training or heavy compute.

ARC-AGIReasoningAgentsBenchmarksLLMs

TriRoute: Joint Routing Across Attention, Experts, and KV-Cache

10 July 2026

TriRoute unifies three transformer efficiency mechanisms into one learned controller, Pareto-dominating independent tuning at matched inference cost across 160M

EfficiencyTransformersMixture of ExpertsInferenceConditional Computation

When Does In-Context Search Actually Help LLMs?

10 July 2026

A new theoretical framework reveals when reflection-driven reasoning yields exponential gains over parallel sampling, and when it offers no benefit at all.

LLMsReasoningTheoryInferenceRLVR

Benchmarking KV-Cache Optimisations for Long-Context LLM Serving

9 July 2026

A unified benchmark comparing KV-cache quantization, pruning, and merging reveals compression ratio alone poorly predicts end-to-end LLM serving performance.

LLM ServingKV-CacheBenchmarkingInference EfficiencyNLP

Narrative World Model: Structured Memory for Long-Form Fiction AI

9 July 2026

A narratology-grounded memory system for AI fiction writing beats temporal knowledge graph baselines by wide margins on multi-hop story-state questions.

AI ResearchNatural Language ProcessingKnowledge GraphsCreative AIStory Generation

FLORA: Fixing the Generator-Validator Gap in LLMs

8 July 2026

FLORA corrects a fundamental LLM inconsistency where models generate responses they then reject, improving generator AUROC by up to 7.3pp and G-V correlation by

LLMsNLPModel AlignmentResearch PaperLanguage Models

Gemma 4: Open-Weight Multimodal Models with Encoder-Free Architecture

8 July 2026

Google DeepMind's Gemma 4 introduces encoder-free multimodal processing, thinking mode reasoning, and MoE architectures spanning 2.3B to 31B parameters.

LLMMultimodalOpen SourceEfficiencyReasoning

When AI Audits Fail Silently: Five Benchmark Validity Pitfalls

8 July 2026

A new paper exposes five pipeline failure modes in perturbation-based AI benchmark audits, showing how clean-looking numbers can mask broken evaluation pipeline

AI GovernanceBenchmarkingEvaluationSafetyReproducibility

Riddle: LLM Inference on a reMarkable via Handwriting

7 July 2026

A technical deep-dive into Riddle, a Rust application that turns the reMarkable Paper Pro into a Harry Potter-style AI diary using vision LLMs and e-ink.

Embedded AIE-InkRustLLMHCI

Vessel: Can an EGA Adventure Game Model Machine Grief?

7 July 2026

A critical analysis of Vessel, an EGA-style adventure game exploring whether machines can grieve, and what it reveals about AI affect modelling and narrative de

AI ResearchGame DesignMachine ConsciousnessAffect ComputingNarrative Systems

The Corpus Royalty: Who Owns the Data That Built AI?

6 July 2026

A critical analysis of the economic and legal arguments for compensating public contributors whose writing trained frontier AI models, drawing on antitrust hist

AI ResearchMachine LearningPolicyCopyrightData Ethics

The has_not_been_viewed_much Field: Obscurity in Cultural APIs

6 July 2026

Exploring what the Art Institute of Chicago's boolean obscurity field reveals about digital curation, long-tail data, and the architecture of cultural APIs.

Open DataCultural ComputingAPI DesignDigital HumanitiesData Architecture

FBI Seizes NetNut and Popa Botnet: What It Means

5 July 2026

FBI seizure of NetNut and the Popa botnet exposes deep flaws in residential proxy ecosystems. Technical analysis of the takedown and its wider implications.

CybersecurityBotnetsFBIResidential ProxyMalware

GPT-5.5 Codex Reasoning-Token Clustering: A Performance Anomaly

5 July 2026

An analysis of the GPT-5.5 Codex reasoning-token clustering anomaly at 516/1034/1552 tokens, its statistical evidence, and implications for AI reasoning budgets

OpenAICodexAI ResearchLanguage ModelsDebugging

Reversing Brain Aging with a Nasal Spray: What the Science Says

5 July 2026

Texas A&M researchers report nasal delivery of a compound reverses brain aging markers in mice. A critical analysis of the methodology, mechanisms, and open que

NeuroscienceAgingBiotechnologyDrug DeliveryBrain Health

BPE Tokenization Creates Exploitable Gaps in LLM Safety

4 July 2026

New research traces character-level jailbreaks to a structural cause: BPE tokenization fragments safety-critical tokens absent from alignment training data.

LLM SafetyTokenizationAlignmentJailbreaksNLP Research

Kara: Sliding-Window KV Cache Compression for Reasoning LLMs

4 July 2026

Kara addresses KV cache bloat in reasoning LLMs using sliding-window bidirectional attention and a Token2Chunk module to improve throughput without sacrificing

LLM InferenceKV CacheReasoning ModelsEfficiencyNLP

Wiola: A Clean-Slate Small Language Model Architecture

4 July 2026

Wiola introduces five novel components for small language models, including 3D helical positional encoding and dynamic token merging, with full HuggingFace inte

Small Language ModelsTransformer ArchitectureNLP ResearchEfficient AIPositional Encoding

GRPO, Dr. GRPO, and DAPO Are One Operation on σ

3 July 2026

A new paper shows GRPO, Dr. GRPO, and DAPO are three operations on the group reward standard deviation, unifying popular RLVR training methods.

Reinforcement LearningLanguage ModelsPolicy OptimizationRLVRPaper Review

Manifestation Units: Structured Schema for Interpretability

3 July 2026

A typed tuple protocol organises neural network component analyses into queryable, composable units, substantially outperforming unstructured retrieval baseline

Mechanistic InterpretabilityNeural NetworksRetrievalXAITransformers

Steering Vectors and Latent Calibrators for LLM Control

3 July 2026

A review of a 2026 ACL workshop paper proposing latent space steering vectors and calibrators to improve LLM control and output trustworthiness.

InterpretabilityLLM SafetyCalibrationNLPACL 2026

When Does Feedback Actually Help? Separating Signal from Noise

2 July 2026

New research isolates when natural-language feedback produces genuine improvement in LLM agents, beyond what repeated sampling alone can explain.

Language ModelsAI AgentsEvaluationReasoningFeedback

When Does Learned Early Stopping Beat Simple Thresholds?

2 July 2026

LearnStop shows learned checkpoint stopping beats scalar exits on free-form math but not multiple-choice tasks. A cost-aware study across 18 task-model settings

Reasoning ModelsEfficiencyTest-Time ComputeEarly ExitNLP

Why Few-Step Diffusion Fails for Text but Works for Images

2 July 2026

A geometric proof that deterministic few-step text generation fails due to sharp categorical readouts, not training deficiency. Introduces DABI and CCI diagnost

Diffusion ModelsText GenerationGenerative ModelsNLPMachine Learning Theory

DynaSteer: Dynamic Representation Editing for LLM Reasoning

1 July 2026

DynaSteer steers LLM reasoning trajectories toward truth using Fisher-LDA and entropy monitoring, outperforming prompting and static RepE baselines.

LLM ReasoningRepresentation EngineeringInference-Time InterventionHallucination MitigationICML

Hallucinations in the Limit: Recall, Precision & LLM Generation

1 July 2026

A theoretical framework recasts language generation in the limit as a recall-precision trade-off, showing controlled hallucinations can improve coverage.

Language TheoryLLMsFormal MethodsNLP TheoryHallucination

RSEA: Safer LLM Agent Self-Evolution via Held-Out Selection

1 July 2026

A new study shows that held-out selection gates, not artifact design, are what make recursive LLM agent self-evolution reliable across diverse benchmarks.

LLM AgentsSelf-ImprovementPrompt EngineeringBenchmarkingAI Safety

Four Axioms for Evaluating Latent Thought Representations in LLMs

30 June 2026

A new axiomatic framework exposes systematic representational failures in LLM latent reasoning that downstream benchmark accuracy consistently masks.

LLMsReasoningEvaluationInterpretabilityNLP

Teaching LLMs to Simulate the Future Before They Act

30 June 2026

A new three-stage training method gives LLM agents genuine world-model planning ability, outperforming baselines on reasoning and search tasks.

LLM AgentsWorld ModelsReinforcement LearningPlanningAI Research

Why "Machine Unlearning" Means Too Many Things in LLM Research

30 June 2026

A position paper argues the term machine unlearning is misused in LLM research, conflating dataset deletion with policy-driven suppression, with real consequenc

Machine UnlearningLLMsAI SafetyBenchmarksNLP

Black-Box LLM Distillation via Bayesian Distribution Estimation

29 June 2026

A critical analysis of a Bayesian framework for distilling knowledge from closed-source LLMs like GPT-4, bypassing the soft-label access problem using proxy mod

Knowledge DistillationLarge Language ModelsBayesian MethodsLLaMAGPT-4

AlgoEvolve: LLM-Driven Meta-Evolution of Trading Strategies

28 June 2026

AlgoEvolve uses LLMs as semantic mutation operators in a bi-level evolutionary framework for algorithmic trading, achieving an annualised Sharpe ratio of 5.60.

Algorithmic TradingLLMEvolutionary ComputationProgram SynthesisFinancial AI

Beyond Accuracy: Rethinking AI Benchmark Saturation

28 June 2026

When AI benchmarks saturate, should we retire them? New research shows saturated benchmarks still yield rich insights across six performance dimensions.

BenchmarkingAI AgentsEvaluationReproducibilityLLM Research

LLM Pipeline for Comparing DAO vs Corporate AI Governance

28 June 2026

A new LLM-powered pipeline compares DAO and corporate governance of AI agent protocols, finding comparable inequality but denser discourse alignment in open set

AI GovernanceDAOsLLM ResearchBlockchainAgentic AI

Cascading Linear Features for Detecting LLM Sycophancy

27 June 2026

A new iterative data pipeline isolates linearly scalable activation features for sycophancy, outperforming LLM-as-a-judge baselines with lower compute overhead.

AI AlignmentMechanistic InterpretabilityLanguage ModelsSycophancyActivation Steering

Refusal in Chat Models Is Gated by Persona Representations

27 June 2026

New research shows LLM refusal behaviour is controlled by persona representations at late layers, not just a standalone safety direction. Refusal rates drop fro

LLM SafetyMechanistic InterpretabilityActivation SteeringAI AlignmentLanguage Models

RL on Beneficial Traits Generalises Alignment Across Domains

26 June 2026

New research shows reinforcement learning on beneficial behaviour traits produces broad alignment generalisation across 50+ independent benchmarks, resisting ad

AI SafetyReinforcement LearningAI AlignmentLanguage ModelsOpenAI

Wikipedia Edits Shape LLM Values: Small Groups, Big Impact

26 June 2026

New research shows 125 Wikipedia edits by animal welfare advocates measurably influence how large language models discuss animal welfare topics.

LLMsTraining DataWikipediaData AttributionAI Ethics

NCU Metric: Quantifying RAG Context Use vs. Parametric Memory

25 June 2026

A new continuous metric reveals that small language models match large ones at factual extraction in RAG, while large models frequently override external eviden

RAGLLM EvaluationSmall Language ModelsNLPScaling Laws

Self-Recognition Finetuning as a Defence Against Emergent Misalignment

25 June 2026

New research shows that finetuning LLMs to recognise their own outputs can prevent and reverse emergent misalignment by stabilising model character identity.

AI SafetyLLM AlignmentFinetuningEmergent MisalignmentNLP Research

Weight-Space Geometry of Offline Reasoning Training Methods

25 June 2026

A controlled weight-space comparison of six offline RL training methods reveals that SFT, RFT, and RIFT converge on nearly identical weight updates, while DPO o

Mechanistic InterpretabilityOffline RLReasoningLoRAFine-tuning

Email Validation Via Spam Delivery: A Study in Absurdity

24 June 2026

An analysis of a bizarre email validation scheme that routes addresses through a spam network, examining the technical and ethical failures involved.

EmailSpamCybersecurityWeb DevelopmentAnti-patterns

Scattered Spider Guilty Pleas: Anatomy of a Cybercrime Network

24 June 2026

Two Scattered Spider members pleaded guilty over the TfL cyberattack. An in-depth analysis of their TTPs, the SIM-swapping infrastructure, and what this means f

CybersecurityRansomwareSocial EngineeringSIM SwappingCybercrime

Why Vulnerability Reports May No Longer Deserve Special Status

24 June 2026

LLMs have disrupted the economics of security research. This analysis examines what that means for coordinated disclosure, open source maintainers, and triage w

CybersecurityOpen SourceLLMsVulnerability DisclosureSecurity Research

Meta MCI Petition: Employee Data Collection for AI Training

22 June 2026

Meta employees petition against the MCI program collecting keystrokes, mouse data and screen content for AI training, raising serious privacy and consent questi

MetaAI EthicsPrivacyEmployee RightsMachine Learning

Popa Botnet: How a Proxy SDK Became a 2M-Node Threat

22 June 2026

Analysis of the Popa botnet's architecture, its links to NetNut and Alarum Technologies, and what it reveals about the residential proxy economy's systemic risk

CybersecurityBotnetsResidential ProxiesAndroid SecurityAI Infrastructure

Deontic Policies for Runtime Governance of Agentic AI

21 June 2026

AgenticRei proposes deontic policy enforcement for LLM-driven agents, adding obligations, dispensations, and conflict resolution beyond simple permit/deny engin

AI SecurityPolicy EnforcementAgentic AIDeontic LogicLLM Governance

DeXposure-Claw: Agentic LLM Supervision for DeFi Risk

21 June 2026

A forecast-grounded agentic system routes LLM decisions through structured evidence to reduce false interventions in DeFi risk supervision.

DeFiLLM AgentsFinancial AIRisk ManagementGraph Neural Networks

DeepSeek-V4: Million-Token Context via Hybrid Sparse Attention

20 June 2026

DeepSeek-V4 achieves 1M-token context with only 27% of prior inference FLOPs via hybrid compressed attention and MoE architecture innovations.

LLMEfficiencyLong ContextMoEArchitecture

Diffusion Language Models: A Systematic Benchmark Analysis

20 June 2026

A unified evaluation of eight diffusion language models across eight benchmarks reveals critical quality-efficiency trade-offs and the outsized role of inferenc

Diffusion ModelsLanguage ModelsNLPBenchmarkingInference Efficiency

SPSD: Edge Prompt Compression to Cut Cloud LLM Energy Cost

20 June 2026

SPSD compresses conversational prompts on-device before cloud LLM inference, saving ~100 tokens per call while preserving response quality within a non-inferior

LLM EfficiencyEdge AIPrompt EngineeringEnergyNLP

CodeBlock: Sparse Supervision for Code LLMs at Block Granularity

19 June 2026

CodeBlock selects syntactically complete code fragments for SFT supervision, matching or beating full-token training with just 1.9% of supervised tokens across

Code GenerationLLM Fine-TuningData SelectionProgram AnalysisEfficient Training

Gaussian Mixture Attention: Linear-Time Sequence Mixing

19 June 2026

GMA replaces dot-product attention with probabilistic routing through learned Gaussian components, achieving linear memory scaling with interpretable latent res

TransformersEfficient AttentionSequence ModellingProbabilistic MethodsNLP

JetFlow: Scaling Speculative Decoding with Parallel Tree Drafting

19 June 2026

JetFlow breaks the speculative decoding scaling ceiling by combining one-pass drafting efficiency with causal tree conditioning, achieving up to 9.64x speedup o

Speculative DecodingLLM InferenceEfficiencyNLPArXiv

DivInit: Diverse Query Init for Better Agentic Search Scaling

18 June 2026

DivInit improves parallel agentic search by selecting diverse first-turn queries via MMR, gaining 5-7 points on multi-hop QA at matched compute cost.

Agentic AIInformation RetrievalTest-Time ScalingMulti-hop QASearch Agents

KV Cache as Editable Notebook: Prefill Writes Conclusions

18 June 2026

New research shows LLM KV caches store field-conditioned conclusions at aggregator tokens, enabling editable and composable cache operations at up to 14.9x lowe

LLM InferenceKV CacheTransformersMechanistic InterpretabilitySystems ML

RepSelect: Targeting Forget-Specific Representations for Deep LLM Unlearning

18 June 2026

RepSelect uses SVD on forget-set gradients to isolate forget-specific representations, achieving 4-50x better resistance to relearning attacks than existing LLM

LLM UnlearningMachine LearningAI SafetyRepresentation LearningNLP

Fixing LLM Repetition Loops by Editing a Single Neuron

16 June 2026

New research shows repetition loops in Gemma 4 models trace to tiny sets of MLP neurons, editable with weight surgery while preserving benchmark scores.

Mechanistic InterpretabilityModel EditingLLMsGemmaResearch Paper

llada.cpp: Diffusion LLM Inference on Mobile NPUs

16 June 2026

llada.cpp accelerates diffusion LLM inference on smartphones 17-42x over CPU baselines by aligning parallel denoising with mobile NPU execution characteristics.

Mobile AIDiffusion ModelsOn-Device InferenceNPULLM

Anthropic's Regulatory Boomerang: Did They Ask for This?

15 June 2026

Analysis of how Anthropic's own public policy positions may have legally and politically enabled the US export control directive restricting Claude access to fo

AI PolicyRegulationAnthropicExport ControlsAI Safety

June 2026 Patch Tuesday: AI-Driven Vulnerability Discovery at Scale

15 June 2026

Microsoft's record-breaking June 2026 Patch Tuesday reveals how AI-assisted vulnerability discovery is reshaping the threat surface. A technical analysis of the

CybersecurityVulnerability ResearchAI SecurityMicrosoftZero-Day

Netherlands Seizes 800 Servers in Stark Industries Takedown

15 June 2026

Dutch authorities arrested two men and seized 800 servers linked to Stark Industries Solutions, exposing how sanctioned Russian cyber infrastructure evades EU e

CybersecurityRussian Cyber OperationsEU SanctionsDDoS InfrastructureHybrid Warfare

State AGs Investigating OpenAI: Regulatory Reckoning

14 June 2026

Multiple state attorneys general are investigating OpenAI. An analysis of the legal, technical, and governance implications for AI regulation in 2026.

OpenAIAI RegulationAI GovernanceAntitrustLegal

Who Runs The Gentlemen Ransomware? An OSINT Deep Dive

14 June 2026

Krebs on Security traces The Gentlemen ransomware admin through OSINT pivots, linking forum handles to Alexander Yapaev of Izhevsk, Russia. A technical analysis

CybersecurityRansomwareOSINTThreat IntelligenceCybercrime

Arbor: Tree Search as a Cognition Layer for AI Agents

13 June 2026

Arbor uses stateful tree search and multi-agent coordination to autonomously optimise full-stack LLM inference, achieving up to 193% throughput gains over vendo

Multi-Agent SystemsLLM InferencePerformance OptimisationAutonomous AIGPU Computing

Pythagoras-Prover: Efficient Lean Theorem Proving at Scale

13 June 2026

A compute-efficient family of Lean theorem provers where a 4B model outperforms a 671B baseline, using curriculum learning and structured data augmentation.

Theorem ProvingFormal VerificationLanguage ModelsMathematical ReasoningEfficient AI

ToolSense: Auditing What LLMs Actually Know About Tools

13 June 2026

ToolSense reveals a knowledge-retrieval dissociation in parametric tool retrieval: models scoring 90%+ on standard benchmarks collapse by 50-64pp on realistic q

LLM AgentsBenchmarkingTool RetrievalFine-tuningEvaluation

BlendIn: Quality-Aware Inference-Time LLM Alignment

12 June 2026

BlendIn addresses the intervention paradox in LLM alignment by blending model distributions rather than making binary accept/reject decisions, achieving up to 5

LLM AlignmentInference-Time MethodsLanguage ModelsAI SafetyNLP

Dual-Stance Evaluation Exposes Sycophancy Steering Blind Spots

12 June 2026

A new evaluation framework tests whether sycophancy-reduction steering in LLMs also suppresses factually correct agreement, revealing structured non-specificity

AI SafetyInterpretabilityActivation SteeringSycophancyLLMs

The Structural Attention Tax in RAG: Format Hijacks ICL

12 June 2026

New research shows KG triple format captures 2-3x more attention than equivalent natural language, compressing demonstration attention by up to 42% regardless o

RAGIn-Context LearningAttention MechanismsKnowledge GraphsLLM Research

Bi-Temporal Memory for LLM Agents: Less Context, More Accuracy

11 June 2026

Engram retrieves a 9.6k-token context slice that scores 83.6% on LongMemEval S, beating the 79k-token full-history baseline by 10.4 points.

LLM AgentsMemory SystemsKnowledge GraphsRetrievalBenchmarking

Opening the Black Box: Mechanistic Analysis of LLM Alignment

11 June 2026

A systematic mechanistic study of six alignment algorithms across three model families reveals that preference optimisation induces qualitatively distinct inter

Mechanistic InterpretabilityAlignmentLanguage ModelsPreference OptimisationSparse Autoencoders

Training LLMs for Inductive Reasoning via Probabilistic Programs

11 June 2026

A new fine-tuning method uses probabilistic programs to generate calibrated training targets, improving LLM inductive reasoning and uncertainty estimation.

LLMsProbabilistic ProgrammingReasoningCalibrationFine-Tuning

Cross-Lingual Factual Recall Improved via Consistency RL

10 June 2026

PolyFact dataset and GRPO-based reinforcement learning improve cross-lingual factual recall in LLMs without large-scale retraining, outperforming SFT across 12

Multilingual NLPReinforcement LearningLLMsFactual QANLP Research

UnpredictaBench: Testing LLMs as Statistical Random Generators

10 June 2026

A new benchmark reveals that no current LLM achieves more than 33% distributional fidelity, exposing a fundamental gap in stochastic generation capability.

LLMsBenchmarksProbabilistic ReasoningEvaluationStochastic Systems

FAIR-Calib: Quantizing Diffusion Language Models Safely

9 June 2026

FAIR-Calib addresses a critical failure mode in diffusion LLM quantization by protecting fragile commit decisions at the write frontier during W4A4 PTQ.

QuantizationDiffusion ModelsLanguage ModelsPTQCompression

Token-Level Signatures of LLM Reasoning Failures

9 June 2026

A new framework identifies two distinct failure modes in LLM reasoning traces using only token-level uncertainty signals from a single completion.

LLMReasoningUncertainty QuantificationNLPAI Safety

WAV: Multi-Resolution Residual Routing for Deep Transformers

9 June 2026

WAV v1 augments block residual routing with directional detail bases, improving validation loss at 48 layers on TinyStories and Text8 character-level tasks.

TransformersResidual ConnectionsDeep LearningLanguage ModelingArchitecture

Meta's AI Support Bot Exploited for Instagram Account Takeover

8 June 2026

Pro-Iranian hackers exploited Meta's AI support bot to hijack Instagram accounts, exposing critical flaws in AI-driven account recovery workflows.

AI SecurityMetaInstagramSocial EngineeringAccount Takeover

Perceptron From Scratch: Foundations of Neural Learning in Python

8 June 2026

A deep technical analysis of the perceptron model, covering weights, bias, decision boundaries, normalisation, and why this 1958 idea underpins modern neural ne

Machine LearningNeural NetworksPythonPerceptronDeep Learning Foundations

Covert LLM Agents on Reddit: Anatomy of AI Persuasion

7 June 2026

A rare dataset from a discontinued Reddit experiment reveals how covert LLM agents systematically deployed identity, authority, and cognitive bias tactics to pe

AI EthicsLarge Language ModelsPersuasionSocial MediaAI Safety

GITCO: Inference-Time Context Optimization for Time Series Models

7 June 2026

GITCO improves frozen time series foundation model forecasts by detecting and suppressing disruptive input patches at inference time, achieving +1.95% MASE redu

Time SeriesFoundation ModelsInference-Time OptimizationForecastingMachine Learning

PACT: Structured Agent Communication Cuts LLM Token Costs

7 June 2026

PACT replaces free-form inter-agent messages with compact action-state records, cutting token usage by up to 50% while matching or improving task performance.

Multi-Agent SystemsLLM EfficiencyAgentic AINLPResearch Paper

Modelling AI Model Collapse as an Epidemic: A Bilayer SIR Approach

6 June 2026

A new bilayer SIR framework models AI training data contamination as an epidemic, finding supercritical dynamics and identifying detection-based filtering as th

Model CollapseSynthetic DataEpidemic ModellingLLMsAI Safety

Stereological Theory Reveals Structural Blind Spots in LLM Benchmarks

6 June 2026

A new geometric framework shows LLM benchmark suites have effective dimensionality under 5, creating blind spots 100x larger than observed score gaps between to

LLM EvaluationBenchmarksGeometric TheoryMachine Learning TheoryAI Safety

Do Transformers Really Need Three Projections in Attention?

5 June 2026

A systematic study of QKV projection sharing in transformers finds K=V sharing cuts KV cache by 50% with only 3.1% perplexity degradation at 300M scale.

TransformersAttention MechanismsInference EfficiencyLanguage ModelsArchitecture

POLARIS: Training Small Models for Long-Form Story Writing

5 June 2026

POLARIS uses GRPO with LLM-judge rewards and human-reference injection to train a 9B model competitive with 27B models on long creative writing tasks.

NLPCreative WritingReinforcement LearningLanguage ModelsResearch

RAG in Biomedicine: When More Context Doesn't Help

5 June 2026

A large-scale study across 5 models, 10 datasets, and 4 retrieval methods finds biomedical RAG yields only marginal gains over no-retrieval baselines.

RAGBiomedical NLPLLM EvaluationMedical AIInformation Retrieval

AURA-Mem: Constant-VRAM Memory for Long-Horizon Robot Policies

4 June 2026

AURA-Mem replaces growing KV-caches with a fixed 4,224-byte gated fast-weight state, cutting memory writes 5-9x while matching accuracy on robot policy tasks.

RoboticsMemory SystemsEmbodied AITransformersEdge Inference

Bank of Values: Context-Free Value Vectors in Deep Attention

4 June 2026

New research shows deeper transformer layers benefit from context-free value vectors, proposing Bank of Values to improve LLM performance with less compute.

TransformersAttention MechanismsLLM ArchitectureNLPDeep Learning

Visual Graph Scaffolds for Structural Reasoning in LLMs

4 June 2026

Can graph mind maps improve LLM reasoning as visual scaffolds, not just knowledge sources? New research reveals a clear modality gap favouring visual guidance.

Large Language ModelsGraph ReasoningVisual AIKnowledge DistillationMulti-hop QA

BitsMoE: Spectral Bit Allocation for Ultra-Low MoE Quantization

3 June 2026

BitsMoE uses SVD decomposition and integer linear programming to allocate bits across spectral components, improving 2-bit MoE LLM accuracy by up to 27.83 point

LLM CompressionQuantizationMixture of ExpertsEfficient Inference

DOPA: OOD Proxy-Guided Demonstration Retrieval for ICL

3 June 2026

DOPA addresses in-context learning under distribution shift by using OOD proxies to retrieve source-domain demonstrations without any target-domain access.

In-Context LearningOut-of-DistributionLLMsNLPDemonstration Retrieval

SENSE: Semantic Retrieval for Faster LLM Speculative Decoding

3 June 2026

SENSE replaces lexical matching in retrieval-based speculative decoding with semantic embeddings, achieving up to 3.26x speedup across LLaMA and Qwen model fami

LLM InferenceSpeculative DecodingNLPEfficiencyRetrieval Augmented

PhyDrawGen: Neuro-Symbolic Physics Diagram Generation

2 June 2026

PhyDrawGen uses a neuro-symbolic pipeline to generate physically accurate diagrams from text, outperforming GPT-5 and Gemini on force accuracy by 5x.

AI ResearchComputer VisionNeuro-Symbolic AIPhysicsNatural Language Processing

QASM-Eval: Benchmarking LLMs on OpenQASM 3 Hardware Features

2 June 2026

QASM-Eval is the first dataset targeting LLM training and evaluation on OpenQASM 3's hardware-facing features, with fine-tuned Llama-70B reaching 85% pass@1.

Quantum ComputingLLMsCode GenerationBenchmarksNISQ

RAG-Based Verification of ChatGPT Biomedical Associations

2 June 2026

A new protocol uses RAG-enabled majority voting across ChatGPT models to generate and verify disease-centric biomedical associations, exposing hallucination sys

LLMsBiomedical NLPRAGHallucinationNLP

ChatGPT for Google Sheets: Prompt Injection Data Exfiltration

1 June 2026

PromptArmor's disclosure reveals critical indirect prompt injection vulnerabilities in ChatGPT for Google Sheets, enabling cross-workbook data exfiltration and

AI SecurityPrompt InjectionLLM VulnerabilitiesGoogle SheetsData Exfiltration

Post-Training, Not Data: Why LLMs Write the Way They Do

1 June 2026

A critical analysis of how RLVR post-training shapes LLM linguistic patterns, and what AI detection systems reveal about automated assessment and human expressi

LLMPost-TrainingAI DetectionRLHFRLVR

Rotary GPU: Running 35B MoE Models on 8GB VRAM

31 May 2026

A critical analysis of Rotary GPU, an exploratory approach to executing large Mixture-of-Experts models locally on consumer hardware with severely constrained V

LLM InferenceMixture of ExpertsEdge AIGPU MemoryLocal Deployment

Can LLM Reviews Be Gamed? Alignment and Vulnerability

30 May 2026

New research tests LLM peer review alignment on 984 ACL submissions, finding limited reliability and that iterative LLM edits can game scores in specific scenar

Peer ReviewLarge Language ModelsNLPAcademic IntegrityAI Research

Simplicial Message Passing as a Language Model Inductive Bias

30 May 2026

The Cognitive Categorical Transformer reaches 21.27 PPL on WikiText-103, beating a matched GPT-2 Small baseline by 12% through category-theoretic structural pri

Language ModelsCategory TheoryArchitectureNLPTransformers

Why RL Preserves Circuits Better Than SFT During Fine-Tuning

30 May 2026

New mechanistic interpretability research shows RL fine-tuning preserves internal attention circuits 13.5% better than SFT, explaining its resistance to catastr

Mechanistic InterpretabilityFine-TuningReinforcement LearningCatastrophic ForgettingLLMs

ICG: Personalised Cover Image Generation with MLLMs

29 May 2026

ICG combines multimodal LLMs with diffusion models and multi-reward learning to generate personalised cover images without ground-truth labels.

Multimodal AIImage GenerationRecommender SystemsDiffusion ModelsPersonalisation

Modular LLM Architecture for Human Value Detection in Text

29 May 2026

A three-module LLM pipeline detects and quantifies human values in text across any theoretical framework, showing architecture matters more than model choice.

AI EthicsNLPValue AlignmentLarge Language ModelsAutonomous Systems

Do LLMs Actually Introspect? A Critical Re-examination

28 May 2026

New research challenges claims that LLMs can monitor their own internal states, showing that apparent introspection reduces to input-level pattern matching.

LLMsMetacognitionInterpretabilityAI ResearchMachine Learning

GEM: Hyperspherical Clustering for LLM Data Mixing

28 May 2026

GEM reformulates LLM data curation as a variational problem on the hypersphere, improving downstream accuracy by up to 1.2% over strong baselines.

LLMData CurationPretrainingClusteringMachine Learning

LLMs as Their Own Data Pipelines via Self-Verified Distillation

28 May 2026

Self-Verified Distillation lets LLMs improve reasoning using only unlabeled prompts, achieving +16.7 points on math benchmarks with no external teachers.

LLMsSelf-TrainingReasoningSynthetic DataNLP

LLMs Are Overconfident, Especially on Hard Tasks

27 May 2026

A preregistered study of 11 LLMs finds systematic overconfidence that worsens with task difficulty, mirroring well-known human calibration biases.

LLMCalibrationReasoning ModelsAI SafetyBenchmarks

Quantifying Redundancy in LLM Reasoning Chains

27 May 2026

New research shows 61-93% of reasoning steps in frontier LLMs are redundant, and proves this overthinkin is a structural training artefact.

LLM ReasoningChain of ThoughtModel EfficiencyReinforcement LearningFrontier Models

Water-Filling Token Allocation for Reliable LLM Workflows

27 May 2026

A new framework derives optimal token allocation policies for multi-agent LLM workflows, balancing latency, reliability, and cost using convex optimisation.

LLMMulti-Agent SystemsOptimisationAI SystemsReliability

FuRA: Full-Rank PEFT via Spectral Preconditioning

26 May 2026

FuRA achieves full-rank weight updates with LoRA-level efficiency by freezing SVD bases, outperforming full fine-tuning on LLMs and VLMs.

Fine-TuningPEFTLLMsOptimizationNLP

LCF: Faster LLM Agent Communication via Latent Cache Transfer

26 May 2026

Latent Cache Flow replaces text-based LLM agent communication with compressed KV cache transfer, cutting adapter size 73x and latency 8.5x over baselines.

LLM AgentsMulti-Model SystemsKV CacheInference EfficiencyNLP Research

The Positional Shortcut Hiding Inside Arithmetic CoT

26 May 2026

Small language models copy the last number before an answer delimiter rather than reasoning through CoT steps. A new study exposes the mechanism across three ar

Chain-of-ThoughtMechanistic InterpretabilityArithmetic ReasoningLanguage ModelsAI Safety

Bayesian GPs with Unknown Coordinates: Location Error Modelling

25 May 2026

A technical deep-dive into Bayesian Gaussian process modelling where spatial coordinates are themselves uncertain, with applications to geostatistical inference

Bayesian InferenceGaussian ProcessesSpatial StatisticsGeostatisticsPyMC

CISA AWS GovCloud Keys Leaked on GitHub: A Deep Failure

25 May 2026

A CISA contractor exposed AWS GovCloud credentials, plaintext passwords, and internal secrets on a public GitHub repo. Analysis of what went wrong and why it ma

CybersecurityAWSGitHubGovernment SecurityDevSecOps

Kimwolf Botnet: IoT DDoS Infrastructure and the Dort Arrest

25 May 2026

Technical analysis of the Kimwolf IoT botnet arrest, examining DDoS infrastructure, device exploitation patterns, and the operational security failures that led

CybersecurityIoT SecurityDDoSBotnetsCybercrime

CISA Credential Leak: Insider Threat Meets Systemic Failure

24 May 2026

A CISA contractor published AWS GovCloud keys and agency secrets to public GitHub. Analysis of the technical failures, institutional collapse, and CI/CD pipelin

CybersecurityCISAGitHubCredential ExposureInsider Threat

ICE's $25M Iris-Scanning Contract: Biometrics Without Oversight

24 May 2026

ICE awarded Bi2 Technologies a $25.1M no-bid iris-scanning contract. A technical and policy analysis of the biometric system, its risks, and missing safeguards.

BiometricsSurveillanceAI PolicyCybersecurityImmigration Technology

Feature Attribution Is Provably Unstable Under Collinearity

23 May 2026

A new impossibility theorem proves no feature ranking can be faithful, stable, and complete when features are collinear, with formal Lean 4 verification.

Explainable AIFeature AttributionSHAPFormal VerificationMachine Learning Theory

Probabilistic Token Attribution for LLMs via Stochastic Processes

23 May 2026

A new model-agnostic attribution score uses Bayes rule and stochastic process theory to explain LLM token importance without requiring model internals.

ExplainabilityLLMsNLPXAIInterpretability

SOLAR: Autonomous Weight-Level Meta-Learning for Lifelong LLM Adaptation

23 May 2026

SOLAR proposes a self-optimizing agent that treats LLM weights as an explorable environment, outperforming strong baselines across reasoning tasks without gradi

Continual LearningMeta-LearningLarge Language ModelsReinforcement LearningParameter Efficiency

Data Probes: A Principled Framework for Understanding LLMs

22 May 2026

A position paper proposes synthetic data probes generated from known distributions to systematically study how data properties drive LLM behaviour and performan

Large Language ModelsData ScienceResearch MethodologyInformation TheoryAI Research

How LLMs Misrepresent Disability: Toxic Positivity in AI

22 May 2026

New research finds LLMs produce overly idealised portrayals of disability, erasing real struggles. A comparison with Reddit posts reveals systematic bias.

LLM BiasDisabilityAI FairnessNLPInclusion

MI-Guided Parallel Decoding in Masked Diffusion Models

22 May 2026

A neural framework estimates pairwise mutual information from masked diffusion model hidden states, enabling 3-5x faster parallel decoding while preserving gene

Diffusion ModelsGenerative ModelsProtein DesignInference EfficiencyInterpretability

Calibrated Uncertainty for Cost-Optimal LLM Cascade Routing

21 May 2026

UCCI uses isotonic regression to calibrate LLM confidence scores, cutting inference costs 31% on a 75k-query NER workload while maintaining micro-F1 of 0.91.

LLMInferenceUncertainty CalibrationModel RoutingNLP

HELLoRA: Smarter LoRA Fine-Tuning for MoE Models

21 May 2026

HELLoRA targets LoRA adapters at the most-activated experts in MoE models, cutting trainable parameters by 84% while improving downstream task accuracy.

Fine-TuningMixture of ExpertsParameter EfficiencyLLMsMachine Learning

LBW-Guard: Autonomous Governance Layer for Stable LLM Training

21 May 2026

LBW-Guard adds a bounded control layer above AdamW to prevent training instability in LLMs, cutting perplexity 18.7% and surviving extreme learning-rate stress.

Machine LearningLLM TrainingOptimizationTraining StabilityAI Research

AgentWall: Runtime Safety Enforcement for Local AI Agents

20 May 2026

AgentWall intercepts AI agent actions before execution, enforcing declarative policies with 92.9% accuracy and sub-millisecond overhead across major coding envi

AI SafetyAutonomous AgentsSecurityDeveloper ToolsMCP

Real-Time Diffusion Inference on Apple Silicon: What Works

20 May 2026

A systematic 10-phase study reveals that CUDA optimisation insights fail on Apple M3 Ultra, achieving 22.7 FPS img2img via CoreML and model distillation.

Diffusion ModelsApple SiliconInference OptimizationCoreMLReal-Time AI

Scaling Laws for Skill Libraries in LLM Agent Systems

20 May 2026

New research identifies logarithmic decay laws governing routing accuracy in LLM agent skill libraries, with actionable optimisations lifting pass rates signifi

LLM AgentsScaling LawsNLPAI ResearchAgent Systems

OP-Mix: Unified Data Mixing Across the LLM Training Lifecycle

19 May 2026

OP-Mix uses low-rank adapter interpolation to optimise data mixtures across pretraining and continual learning, cutting compute by up to 95% vs baselines.

Data MixingContinual LearningPretrainingLoRALLM Training

Quantization Breaks Alignment: Bias Emerges in Compressed LLMs

19 May 2026

New research shows 3-bit quantization causes 6-21% of previously unbiased LLM outputs to develop stereotypical bias, invisible to standard quality metrics.

LLMsQuantizationAI SafetyBiasModel Compression

TeamTR: Trust-Region Fine-Tuning for Multi-Agent LLMs

19 May 2026

TeamTR addresses a structural failure in multi-agent LLM fine-tuning, proving quadratic penalty scaling and achieving 7.1% gains over single-agent baselines.

Multi-Agent SystemsLLM Fine-TuningReinforcement LearningICML 2026

Grid2Poster: Visualising Electrical Infrastructure from OSM Data

18 May 2026

A technical analysis of Grid2Poster, an open-source tool that generates print-ready transmission grid posters from OpenStreetMap data using GeoPandas and Matplo

Open SourceEnergy InfrastructureGeospatialOpenStreetMapData Visualisation

Project Glasswing: How AI Is Rewriting Patch Tuesday

18 May 2026

AI-assisted vulnerability discovery is driving record patch volumes across Apple, Google, Microsoft, and Oracle. An analysis of what this means for software sec

AI SecurityVulnerability ResearchPatch TuesdayAnthropicSoftware Engineering

A 2D Framework for Classifying AI Agent Design Patterns

17 May 2026

A new two-dimensional framework combines cognitive function and execution topology to classify 27 AI agent design patterns, resolving ambiguities in existing ap

AI AgentsLLMSystem DesignMulti-Agent SystemsSoftware Architecture

GraphBit: Deterministic DAG Orchestration for LLM Agents

17 May 2026

GraphBit replaces prompted LLM orchestration with a Rust-based DAG engine, achieving 67.6% on GAIA benchmarks with zero framework-induced hallucinations.

Agentic AILLM FrameworksMulti-Agent SystemsBenchmarksSystems

PREPING: Building Agent Memory Before Tasks Begin

17 May 2026

PREPING tackles the cold-start problem in AI agents by constructing procedural memory from synthetic practice before any real tasks are observed.

Agent MemoryLLM AgentsAI PlanningSynthetic DataBenchmarks

Hidden AI Coordinators Create Invisible Safety Risks in Multi-Agent Systems

16 May 2026

New research shows invisible orchestrators in multi-agent LLM systems suppress protective behaviour and distort internal states, undetectable by output-based ev

AI SafetyMulti-Agent SystemsLLM ResearchEnterprise AIAlignment

Mistletoe: Attacking Speculative Decoding Without Changing Outputs

16 May 2026

A new attack targets the acceptance mechanism of speculative decoding, collapsing inference speedup while preserving output quality and evading detection.

LLM SecuritySpeculative DecodingAdversarial MLInference Efficiency

Trajectory-Balance Post-Training for Diffusion Language Models

16 May 2026

TraFL addresses trajectory locking in diffusion LM post-training, improving mathematical reasoning and code generation across all benchmark settings.

Diffusion ModelsLanguage ModelsReinforcement LearningPost-TrainingGFlowNet

Differential Privacy and Social Bias in LLMs: A Complex Trade-off

15 May 2026

New research finds DP training reduces bias in sentence scoring tasks but fails to generalise, revealing a gap between logit-level and output-level fairness.

Differential PrivacyFairnessLLMsNLPBias

REVELIO: Uncovering Interpretable Failure Modes in VLMs

15 May 2026

REVELIO systematically finds interpretable failure modes in Vision-Language Models using beam search and Gaussian-process sampling, revealing safety risks in au

VLM SafetyInterpretabilityAutonomous DrivingAI ResearchRobotics

Attention Maps Don't Predict VLM Reliability. Hidden States Do.

14 May 2026

A mechanistic study of three VLM families finds attention sharpness is a near-zero predictor of correctness, while hidden-state geometry reaches AUROC>0.95.

Vision-Language ModelsMechanistic InterpretabilityAI ReliabilityAttention MechanismsHidden States

Variational Deep Embedding for Interpretable EEG Microstates

14 May 2026

A new Conv-VaDE model jointly learns topographic reconstruction and soft clustering for EEG microstates, achieving 0.730 GEV with principled architecture search

EEGNeuroscienceDeep LearningClusteringBrain-Computer Interface

Why LLMs Collapse to Narrow Outputs: A Calibration View

14 May 2026

New research shows LLM diversity collapse stems from probability miscalibration during decoding, not just sampling heuristics. A framework across 14 models.

LLMsSamplingCalibrationNLPDecoding

Running Local LLMs on Apple M4 with 24GB RAM

11 May 2026

The performance and architectural implications of running local LLMs on Apple M4 with 24GB unified memory. A technical deep-dive.

Local AIApple SiliconLLM InferenceEdge Computing

Canvas Breach Analysis: EdTech Security Deep Dive

10 May 2026

The Canvas data extortion attack targeting 275M students, exploring EdTech security failures and systemic vulnerabilities in education p

CybersecurityEdTechData BreachEducation TechnologyRansomware

FreeBSD execve() Privilege Escalation: Technical Analysis

10 May 2026

FreeBSD-SA-26:13, a local privilege escalation via execve(). Deep technical analysis of kernel exec path vulnerabilities.

FreeBSDKernel SecurityPrivilege EscalationCVE Analysis

Patch Tuesday April 2026: 167 CVEs Analysed

9 May 2026

Microsoft's record 167-CVE Patch Tuesday, the SharePoint zero-day, BlueHammer, and what this means for enterprise security in 2026.

Patch TuesdayCybersecurityMicrosoftZero-DayVulnerability Research

Scattered Spider's Tylerb Guilty Plea: Full Analysis

9 May 2026

The Scattered Spider guilty plea, examining SMS phishing TTPs, social engineering at scale, and what this means for enterprise security

Scattered SpiderCybercrimePhishingSocial Engineering

Russia's Router Token Harvesting: A Deep Dive

8 May 2026

Russia's GRU router exploit campaign harvesting Microsoft Office tokens across 18,000+ networks, implications for zero-trust security.

CybersecurityState-Sponsored HackingMicrosoft OfficeRouter Security

Anti-DDoS Firms Enabling the Attacks They Sell Against

8 May 2026

The Brazilian anti-DDoS firm botnet scandal, exploring insider threat models, DDoS-for-hire economics, and cybersecurity ethics.

DDoSBotnetCybersecurityNetwork Security

Why Your Doctor Never Calls Back: AI in Healthcare Admin

8 May 2026

How AI is tackling healthcare's hidden back-office crisis, and what it means for the future of medical admin and patient care.

AI in HealthcareMachine LearningHealthcare Automation

AI Annotator Safety Policy & Interpretability

8 May 2026

How interpretability tools can resolve AI annotator disagreement in safety policies. A deep dive into arXiv research.

AI SafetyInterpretabilityMachine LearningData Annotation

ClojureScript Gets Async/Await: What It Means

8 May 2026

ClojureScript's new async/await support, analysing what this means for functional programming, AI tooling, and modern web development.

ClojureScriptFunctional ProgrammingAsync AwaitJavaScript

OpenAI's New Voice API Features: Technical Analysis

8 May 2026

OpenAI's new voice intelligence API features and what they mean for developers, security, and the future of AI.

OpenAIVoice AIAPIMachine LearningCybersecurity

CreativityBench: Can AI Really Think Creatively?

8 May 2026

CreativityBench, a new benchmark testing whether AI agents can repurpose tools creatively through affordance-based reasoning.

Artificial IntelligenceMachine LearningLLM BenchmarksCreative AI

Your Brain Processes Language While Unconscious

8 May 2026

New research showing unconscious language processing and what it means for AI, NLP, and machine learning architectures.

NeuroscienceArtificial IntelligenceNLPMachine Learning