rea: Reverse Engineering Apps and Native Binaries with Agents
The GitHub project rea proposes using agents for reverse engineering, covering app behavior down to native binaries, offering a new direction for automated analysis.
01 DAILY RESEARCH SIGNALS / 2026-10-07
Compress market noise
into research-ready signals.
A curated stream of quantitative research, open-source projects, and AI engineering advances—screened, ranked, and structured for faster research decisions.
Browse today's signals ↓
All sources checked. Pipeline operating normally.
02 CURATED INDEX
Historical signal index
Browse by content type or search titles and tags.
INDEXED 15
01 / REP
GitHub
SELECTED SIGNALS 03
Editorial Diagram Design Toolkit for AI Coding Assistants
Introduces editorial-level diagram design for AI coding assistants like Claude Code, Codex, GitHub Copilot, Factory Droid, and Pi, offering 42 diagram types with self-contained HTML and SVG output, independent of common tools like Mermaid.
AnyPS5 Project Attempts Automatic Porting of PS5 Executables
The project is gaining attention on GitHub Trending for proposing automatic porting of PlayStation 5 executables to Linux and Windows; if feasible, it could help reduce manual cross-platform compatibility efforts.
02 / PPR
Papers
SELECTED SIGNALS 10
Modelling Order Flow in Intraday Electricity Markets with State-dependent Hawkes Processes
The study proposes a multivariate state-dependent Hawkes process that uses a Liquidity Stress Index to identify market states and allows state transitions to change intensity parameters. Calibrated on Dutch intraday cross-border hourly order data from January to March 2024, it finds strongly self-exciting, near-critical order flow, with liquidity stress mainly strengthening transaction self-excitation rather than reshaping cross-side structure.
Investing Is Compression: A Three-Term Decomposition of the Kelly Objective
Borrowing a trick from the universal portfolio algorithm, the author decomposes the Kelly objective into a money term, an entropy term, and a divergence term, showing that maximizing growth is equivalent to minimizing the divergence (in bits) between the predicted distribution and the true distribution. This offers a theoretical view of investing as a compression problem and yields two practical byproducts: in backtests, the difference in log growth between two strategies mea
Stock-JEPA: Prior-Anchored Latent Revision Representation Learning in Equity Markets
This study proposes Stock-JEPA, a joint-embedding predictive framework that uses multi-horizon return and risk statistics from a low-complexity financial model as anchors in the target encoder's latent space, and estimates predictable displacement from the anchor via a context-conditioned revision predictor. Its novelty lies in constraining black-box representation learning with structural priors while retaining an interpretable reference; the authors report that on China and
FractAL: Soft Strategy Selection for Batch-Mode Active Learning
To address the challenge of pre-selecting acquisition strategies in batch active learning, this work proposes soft strategy selection and FractAL, which uses influence-function-based attribution and online mirror descent to dynamically allocate batch budget across strategies without retraining. Across 7 benchmarks spanning classification, regression, and genetic perturbation effect prediction, FractAL matches or outperforms all baselines, offering a practical solution for few
FactorBench: A Portfolio-Aware Benchmark for Automated Factor Mining
FactorBench is a portfolio-aware benchmark that compares roughly five thousand mined factors from nine automated factor mining methods across five equity markets. Its novelty lies in a shared data and evaluation contract that connects heterogeneous discovery algorithms to common signal combination and portfolio construction procedures and traces outputs across factor validity, temporal generalization, predictiveness beyond risk and style exposures, and pool distinctness, help
AegisFlow: A Multi-Agent Agentic AI Framework for Autonomous Remediation and Self-Healing in Fragile Data Ecosystems
Presents an LLM-based multi-agent framework with Watchdog and Repair agents and parallel shadow patching, reducing MTTR from 170 to 3.2 minutes across five failure scenarios.
Execution Assumptions Change Short-Horizon Sharpe Rankings: Evidence from a Synthetic Trading Benchmark
This study tests in a synthetic trading benchmark whether changing order-fill assumptions from idealized close prices to stressed settings with latency, spread, participation, and impact alters the short-horizon Sharpe rankings of LLM and classical baselines. It has direct methodological implications for avoiding strategy superiority claims based only on backtest rankings.
Agent Planning Benchmark: Diagnosing LLM Planning Shortfalls and Tool Robustness
This study introduces the Agent Planning Benchmark (APB), which separates planning from execution and uses 4,209 multimodal cases across 22 domains and five settings to reveal systematic weaknesses in long-horizon planning, tool-noise robustness, calibrated refusal, and inference-time refinement across 12 MLLMs, while validating that APB-guided refinement improves downstream execution metrics.
A Functional Representation of Credit Behavior for Probability of Default Modeling
The study treats account type, balance, and credit limit as continuous-time processes to build a functional representation of borrowers, and introduces a relative time framework to align recurring events across borrowers. Its novelty lies in retaining intra-month continuous information rather than monthly aggregates and using functional logistic regression for interpretable default prediction; its usefulness lies in combining near-XGBoost predictive performance with the inter
Adversarial Training for Deep Hedging via Two-Budget Distributionally Robust Optimization
This work proposes WRAP, a framework that combines reweighting and path perturbation adversarial constraints to address market nonstationarity in deep hedging training. Its novelty lies in converting the inner distributional supremum of DRO into a tractable first-order attack, offering a more robust perturbation scheme for training policies.
03 / VID
Videos
SELECTED SIGNALS 02
Hugging Face Xet Storage Layer Enables Incremental Uploads
Hugging Face's Xet storage layer uploads only changed data pieces instead of re-uploading entire files when data changes. For users frequently updating large datasets, this saves bandwidth and time, improving workflow efficiency.
Course on Building AI Assistants with On-Device Memory
A new course from DeepLearning.AI in partnership with Qdrant teaches how to build AI assistants whose memory stays entirely on-device. Its novelty lies in focusing on on-device memory implementation, avoiding remote servers or network connections; it is practically useful for developers concerned with privacy and offline scenarios.