Papers

Research paper proposes Cortex-FL, an agent-based federated learning system using natural language to coordinate multi-party model training without centralizing data.

Research paper introduces Omakase, a parallel execution framework for multi-proposer blockchains that exploits inter-block parallelism to address execution bottlenecks in L1s like Solana and Aptos.

Research paper applies hidden Markov models to Bitcoin on-chain transaction structure (2012–2026) to detect market regimes without price data.

Research paper implements privacy-preserving electricity markets on Hyperledger Besu using AMM, Poseidon2 commitments, and UltraHonk zero-knowledge proofs for price settlement.

Empirical study of 973,581 Bitcoin Lightning channel force-closes (2018–2026) measuring fee costs under old design before Bitcoin Core 31.0 and May 2026 zero-fee specification change.

Research evaluates two Solidity smart contracts for federated learning reward distribution: public quality scores vs. privacy-preserving zero-knowledge proof variant using non-transferable tokens.

Systematization of knowledge (SoK) survey of Tendermint BFT consensus design choices, covering Tendermint Core and CometBFT protocols, ABCI, and proposed research modifications.

Corpus paper annotates 10,000 Arabic Facebook posts (2014–2024) on cryptocurrency into six functional categories: market commentary, education, promotion, warnings, religious framing, other.

Research presents Auriga, an LLM-agent framework for non-invasive throughput optimization in Hyperledger Fabric by jointly tuning parameters and resolving transaction conflicts.

Research presents CuEVM, a GPU-native EVM execution engine that parallelizes transaction processing to accelerate smart contract execution on Ethereum.

Econometric study tests whether Bitcoin's price follows a power-law trend or reflects statistical artifacts from persistent noise in 15 years of weekly data.

Research finds faster blocks (Base Flashblocks) reduce arbitrageur priority-fee costs but decrease liquidity depth and widen spreads; validates theoretical equilibrium model.

HCI paper describes computational tools co-designed with communities to assess environmental and regulatory impacts of a Bitcoin mining facility in New York.

Study evaluates machine-learning methods for detecting rug pulls on Solana tokens within minutes of launch, testing graph neural networks on limited historical data.

Proposes reward-splitting consensus mechanism to stabilize Bitcoin mining in fee-only regimes post-halving, preventing winner-take-all undercutting attacks.

Theoretical analysis shows proof-of-useful-work with matrix multiplication may paradoxically increase total electricity consumption via subsidy effect (Jevons' paradox).

Presents BioBind protocol binding biometric deduplication to zero-knowledge authentication for decentralized identity, securing concurrent enrollment and credential issuance.

Proposes Agent4Crypto, a multi-agent LLM system for autonomous cryptocurrency trading with specialized agents and market-feedback alignment.

DeFSM replaces centralized federated learning server with smart contract and consensus-free agents for Byzantine-robust submodular maximization using Tukey-depth aggregation.

Proposes quantum-resistant notary scheme for blockchain interoperability combining ML-KEM, ML-DSA, and AES-256 with dynamic trust model, handling 270 tx/sec.

Paper proposes zk-NSGA-II protocol for privacy-preserving collaborative multi-objective optimization on blockchain using RISC Zero zkVM.

Ablation study evaluates local interaction statistics and machine learning models for Ethereum phishing account detection.

Framework analyzes Bitcoin and cryptocurrency market cycle regularities between halving events using Google Trends and Wikipedia data.

Proposes M3FT framework combining micro-level and macro-level reasoning for multi-agent cryptocurrency perpetual futures trading.

Wake 5: open-source Python framework for Solidity testing, fuzzing, and static analysis using in-process EVM (revm) and MCP server integration.

Proposes cryptographically signed replication certificates for academic publications to enable verification of research datasets and outputs.

Discusses agentic AI systems for scambaiting to extract threat intelligence from cryptocurrency scammers and fraud networks.

Measures cost of post-quantum cryptographic signatures on Ethereum for IoT data authentication under EIP-7623 calldata floor and Glamsterdam upgrade.

Proposes Neural Realized Measures to learn intraday volatility aggregation rules supervised by forecasting objectives rather than fixed rules.

Weaver: state-decoupled symbolic execution engine for parallel blockchain transaction processing using lightweight symbolic references and lazy state fetches.

GRIDSHIELD: privacy-preserving peer-to-peer energy trading using zero-knowledge proofs and secure multi-party computation to prevent clearing-operator manipulation while enforcing grid constraints.

Anduin: reputation system using on-chain payment histories to rank AI service providers across 16 blockchains, resistant to Sybil attacks on $6.1M in agentic transactions.

Study of miscompilation vulnerabilities in production zkVMs SP1 and Jolt: compiler bugs in rustc and LLVM can cause valid zero-knowledge proofs of incorrect computations.

Empirical study of volatility response to price trends at minute resolution across three cryptocurrencies and US equities, 2020–2026.

TxBench: verified benchmark dataset for smart-contract attack detectors, validating both malicious and legitimate transactions against execution history.

Analysis of server-side power consumption for ERC-4337 bundlers aggregating user operations on EVM networks.

Revisit of Elliptic++ Bitcoin forensics dataset reveals temporal leakage in train-test split and undocumented feature definitions; proposes corrected pipeline for 55 address-level features.

Blockchain-based DAO for access control in open robot swarms with dynamic join/leave, applied to agriculture and disaster response.

GossipSub and DOG protocol vulnerability: adversary with constant nodes can prevent honest nodes from receiving messages, threatening Ethereum's P2P liveness.

Preregistered comparison of zero-shot time-series foundation models versus SARIMA/GBM baselines for 96-hour carbon-intensity forecasting in Nordic zones.

AMLBench benchmark evaluates LLM agents on tracing illicit funds across 24 real Ethereum cases and 5.05M transfer records to automate blockchain AML investigations.

RegionFence proposes adaptive concurrency control mixing optimistic and pessimistic mechanisms for parallel smart-contract execution within blockchain blocks.

Academic audit identifies methodological flaws in prior research on cryptocurrency trend-factor predictability, portfolio construction, and risk-adjusted returns.

Open Inference Market proposes economic mechanism for peer-to-peer AI inference using blockchain-style checkable rewards to utilize idle GPUs.

AIDA registry documents 5,868 AI service outages (2025–2026) correlated with 5M verified MEV transactions on Ethereum and Base, linking tool failures to trading activity.

Empirical study of 48 real cross-chain bridge attacks (Jan 2022–Aug 2026) proposes six trust-violation attack patterns across multi-component architectures.

Taxonomy of economic-state manipulation attacks in DeFi: 276 real incidents and 91 Code4rena findings analyzed to distinguish root causes from observable price distortions.

LightGBM classification framework identifies centralized exchange addresses on Ethereum using 58 on-chain behavioral features derived from Proof-of-Reserves disclosures (3M addresses).

Failure-mode taxonomy for Ethereum smart contracts categorizes observable deviations from intended behavior and ranks failures by impact severity.

Markovian mathematical model of blockchain mining with infinite miners derives convergence of coupled ODEs and explicit traveling-front position using asymptotic analysis.