01 DAILY RESEARCH SIGNALS / 2026-10-09
A curated stream of quantitative research, open-source projects, and AI engineering advances—screened, ranked, and structured for faster research decisions.
All sources checked. Pipeline operating normally.
02 CURATED INDEX
Browse by content type or search titles and tags.
INDEXED 15
SELECTED SIGNALS 03
This tool proposes using agents to unify reverse engineering from app behavior observation down to native binary analysis, chaining tasks that are usually split across dynamic tracing, disassembly, and symbol recovery. Its value lies in reducing context-switching across layers and letting high-level behavioral evidence directly drive low-level binary understanding.
This project offers editorial-grade diagram design for AI coding tools such as Claude Code, Codex, GitHub Copilot, Factory Droid, and Pi, covering 42 diagram types with self-contained HTML and SVG output, while rejecting shadows and what it calls Mermaid slop. Its novelty lies in embedding diagram design rules directly into AI coding workflows, and its usefulness comes from more controllable, self-contained, visually consistent output.
Alibaba has open-sourced a code review tool that combines deterministic pipelines with an LLM Agent. Its novelty lies in merging rule-based review with language model capabilities, validated at Alibaba's internal scale, and supporting OpenAI and Anthropic compatible interfaces. For teams needing automated code review, it offers built-in multi-language rulesets and line-level comments.
SELECTED SIGNALS 12
This paper addresses the lack of a Rockafellar-Uryasev representation for expectiles under Wasserstein ambiguity sets by developing an envelope theorem for worst-case expectiles. It characterizes the worst-case expectile as the unique root of a worst-case expectation with a two-piece affine integrand, enabling direct application of standard Wasserstein duality. Using this result, a multi-period tri-level mean-expectile portfolio problem is reformulated as four parametric line
This study proposes Ranking Prior Alignment, a framework that distills external ranking priors (from domain experts, teacher models, or LLMs) into any scoring model via a temperature-scaled KL divergence loss to mitigate scarce labels, weak features, and minimal capacity in cold-start credit scoring. Its novelty lies in being model-agnostic and requiring no external model at inference, and it identifies an inverse-scaling pattern where benefits grow as data abundance, model c
This work proposes LLM-BlockFE, a framework that offline-converts long text into executable feature programs so that online inference does not require LLM calls. Its value lies in using long-text information while supporting industrial risk-control deployment, with AUC improvements reported on public and private datasets.
This study systematically benchmarks quantum neural network (QNN) component configurations for financial time series forecasting. Its novelty lies in extending QNN configuration grid search from classification to regression, using the GBP/USD exchange rate as a case study with 1,368 configurations. It provides practical architectural guidance and quantifies the impact of real quantum hardware noise on predictive performance.
This study formulates an optimal trading problem where the trading signal follows an Ornstein-Uhlenbeck process with regime-switching parameters, framing it as a partially observable Markov decision problem. Its novelty lies in comparing three recurrent-neural-network-plus-reinforcement-learning approaches and finding that supplying posterior regime probability estimates enhances cumulative rewards and strategy interpretability.
This work proposes Uncertainty-Aware Budget Allocation (UAB), which estimates difficulty from vote entropy of initial samples and reallocates a fixed generation budget toward questions with high answer disagreement. Compared with uniform allocation, UAB improves average accuracy by 2.3% and up to 5.5% on individual benchmarks across five open-weight models and five reasoning benchmarks, with the largest gains in low-resource settings and no extra model calls.
This study formalizes Adversarial Heuristic Learning (AHL), where AI agents iteratively revise game strategies and supporting software while model weights stay fixed, evaluated on AAArena, a benchmark of 12 authentic adversarial games and 1,920 archived human programs. Its novelty lies in treating agents as learning engines rather than relying solely on reinforcement learning; it is useful because it shows opponent selection and dense feedback can improve policies, yet half o
The first multi-horizon benchmark for second-hand electronics price forecasting, evaluating 11 statistical and deep learning models across six horizons from 1 to 365 days. N-BEATS outperforms statistical baselines at long horizons and a single model generalizes across horizons, offering methodological reference for subscription platform pricing.
This paper reports a 15-day paper-traded autonomous memecoin trading deployment on Solana DEXs, designed to measure three microstructure questions: time-of-day return patterns on a 24/7 venue, net contribution of decision-time filter stacks relative to counterfactual returns of rejected tokens, and structural robustness of small-sample cumulative-return statistics on a heavy-tailed venue. Its main novelty is the provision of public measurement infrastructure and a reproductio
Based on a continually evolving human-directed, agent-engineered quantitative trading system, the study distinguishes outcome evidence, evaluated-object identity, and causal explanation, arguing that external feedback constrains outcome claims while diagnosis resolution depends on provenance and additional identification structure; the proposed prospective binding protocol is not yet validated.
This dataset publicly releases 6,660 algorithmically-rejected trading events from a live Solana decentralized-exchange filter stack, observed continuously over 22 calendar days. It includes 169,123 forward-outcome observations and 1,837 lifecycle snapshots, providing post-rejection price-and-liquidity sample series but no outcome-label column, useful for studying the relationship between filter rejection decisions and subsequent market behavior.
This work formulates discrete choice model specification as a sequential decision-making problem and uses a Deep Q-Network to learn candidate model construction strategies. Its novelty lies in guiding model exploration with a customizable reward function, reducing manual trial-and-error effort and offering a testable path toward automated specification.