01 DAILY RESEARCH SIGNALS / 2026-10-27
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 04
This release fixes an issue where unpadded dates from the model could read future filings, adjusts decision scoring to avoid in-progress bars and settle after long market closures, and corrects portfolio matching, price precision, SEC statement alignment, and other defects to improve backtest and data-reading accuracy.
Openbot is a local-first desktop workspace project on GitHub. Its novelty lies in placing Codex, Claude, and Grok in dedicated workspaces and connecting them through task queues, file sharing, browser control, and agent-to-agent collaboration to form a persistent AI team. It is relevant to researchers and developers who need multi-model coordination with a local-first emphasis.
Postiz is an open-source agentic social media scheduling tool that automates scheduling via AI agents. Its novelty lies in introducing agentic automation to social media management, offering a self-hostable alternative for the open-source community.
The project appears on GitHub Trending, positioned as lifelong personalized tutoring with a dedicated site; the input lacks model, data and evaluation details, so its novelty lies in extending tutoring beyond one-off interactions.
SELECTED SIGNALS 08
Using an NVDA equity and options panel, the study compares stratified and liquidity-weighted PPS sampling for implied volatility surface reconstruction, tests PPS, Neyman and simple random sampling for LSTM training efficiency, and evaluates predictability of prices and forward IV along with activity-clock effects on return distributions.
The study tests whether cross-bond neural networks beat a random walk under strict temporal splitting and dependence-aware inference, finding that small and large price-change MLP ensembles do not outperform persistence in next-day price forecasting, and that direct price-level training degrades performance substantially.
This study proposes a decision audit method to separate reward misspecification from learning error, quantifying whether a learned decision truly implements its stated financial objective rather than merely optimizing a reward value.
Introduces a synthetic spoken-dialogue pipeline that provides earliest-actionable-point supervision and trains an omni-modal language model for chunk-level causal decisions, enabling tool calls before the user finishes speaking.
GraphDancer is a reinforcement learning framework that trains LLMs to interleave reasoning with function calls over heterogeneous graphs, using a graph-aware curriculum that schedules questions by structural difficulty. Trained only on academic and biomedical graphs, it outperforms same-backbone graph-agent baselines on held-out and unseen e-commerce, literature, and legal graphs; the 14B variant also surpasses baselines powered by GPT-4.1-mini and GPT-5-nano on average acros
This study examines whether weekend cryptocurrency volatility predicts Monday variance in equity, commodity, and fixed-income markets, finding that predictive content strengthens significantly after 2024, with a common factor outperforming coin-specific information, though the mechanism remains unresolved.
This study presents the first scalable framework for structural attacks and adversarial training on multigraph neural networks, optimizing structural perturbations via node-edge incidence matrices and introducing a cycle-based regularizer to measure attack plausibility. Experiments show that learned attacks significantly degrade anti-money laundering detection performance, while adversarial training effectively improves robustness.
This study presents an end-to-end image pipeline for detecting partially occluded sugar beets in loading scenarios, introduces a dedicated benchmark and an automatic annotation strategy using SAM 3, compares models trained on manual versus automatically generated labels, and benchmarks multiple inference configurations including edge devices under realistic constraints of precision, inference time, and power consumption.
SELECTED SIGNALS 03
This is a conference promotional short announcing that AI Dev NYC will take place on November 30–December 1, 2026 at Chelsea Piers, NYC, and listing speakers and topic directions. Its value lies in gathering engineering-oriented topics—from model capabilities, synthetic data, and harness optimization to knowledge graphs, agent memory, agent harnesses, self-improving agents, inference, and security—offering agenda reference for practitioners focused on AI engineering implement
A DeepLearning.AI course built with Qodo presents seven AI code review practices with before-and-after examples, arguing that the context given to the reviewer matters more than which model is used. It helps teams systematically check for missed requirements and security issues as AI-generated code increases, and guides building a custom review agent.
This item reports on the QuantInsti EPAT 2026 online convocation, covering graduate numbers and distribution, guest remarks on learning from crises and risk-taking mistakes, and top scorers and projects. Its value lies in community and educational information rather than verifiable market conclusions.