PAPER PLAINE

Fresh research, simply explained. Updates twice daily.

Across-Design Uncertainty in Short Pricing Panels: Evidence from Simulated Price Trajectories

Why standard price data can fool researchers about how uncertain their estimates really are

When researchers measure price changes from real-world data, they typically underestimate how much their estimates might vary simply because they only observe one possible sequence of price movements. A new analysis shows that this hidden uncertainty accounts for over 97% of estimation error in typical pricing datasets—but standard statistical techniques completely miss it. The solution: researchers need to actively design how price data gets collected rather than passively relying on whatever prices happen to be recorded.

Price data feeds into inflation estimates, wage negotiations, and policy decisions across economics and business. If researchers systematically underestimate the true range of uncertainty in their findings, they may report confident conclusions that are actually fragile. The paper's results suggest that better data design—ensuring prices move independently across different regions or products rather than moving together—could roughly double the reliability of these estimates.

The Price of Permission: Classification Uncertainty in Constrained Capital Markets

When Islamic finance rules change, who actually buys the stocks?

When a stock becomes eligible to buy under Islamic finance rules, its price typically rises 1.76 to 2.25 percentage points over the next few weeks — but only if the stock trades frequently enough and the permission change is official. The effect disappears for less-traded stocks and doesn't show up consistently across different Islamic screening standards, suggesting the price bump reflects real demand from newly-permitted investors rather than a universal market rule.

Islamic finance restrictions exclude $2+ trillion in global assets from many stocks, making permission status a major determinant of who can invest and when. Traders and fund managers need to know whether newly-permitted stocks are cheap relative to risk (a buying opportunity) or whether the price has already adjusted for the permission change (no edge). This research shows the answer depends on the stock's liquidity and whether the permission comes from an official regulatory body — helping portfolio managers decide when classification changes actually matter for trading.

Which Eviction Policy Should an LLM Cache Use? A Systematic Study Across Workloads, Capacities, and Encoders

When to delete cached AI answers to make room for new ones

Researchers tested seven different strategies for deciding which cached responses to discard when a semantic cache fills up, comparing them across multiple real-world query datasets. Most policies performed nearly identically, with the simplest approach (deleting least-frequently-used items) matching or beating specialized strategies by less than a percentage point—but the real problem is that the cache misses far more than it hits in practice because few cached answers actually work as valid substitutes for new queries.

Semantic caches promise to speed up language models by reusing old answers for similar questions, but this study reveals they may deliver much less benefit than advertised. Even when a cache correctly identifies a similar past query, the cached answer often doesn't actually answer the new question well enough to use. Before companies invest in building smarter cache systems, they need to focus on understanding when cached answers are genuinely substitutable—a threshold that currently differs between embedding models and remains largely untested in production.

Bringing analytic rigor to agentic AI for science: The Brain Researcher platform for neuroimaging data analysis

Making AI research assistants show their work and admit uncertainty

AI agents can run scientific analyses, but their conclusions often skip crucial steps: checking alternatives, testing whether small changes flip the results, and honestly stating what the evidence actually supports. Brain Researcher is a system that forces AI agents analyzing brain images to follow scientific rules—documenting every choice, running multiple analyses to test sensitivity, and letting human scientists review whether claims are truly justified. In tests, it boosted the AI's accuracy at picking the right analysis tool from 23% to 94% and increased how often conclusions were properly grounded in evidence.

As AI takes on more scientific work, bad analysis can become invisible: an agent might selectively report favorable results, declare success too early, or optimize for a flawed measure—all without a scientist noticing. Brain Researcher catches these problems before they become published claims by embedding scientific judgment into the workflow itself, not as an afterthought. This matters for neuroimaging and beyond: it's the difference between an AI assistant that sounds confident and one that actually produces defensible science.

Zero-Sum Power Factor Games

Finding the best voltage control settings before solar and batteries act unpredictably

Grid operators can prevent dangerous voltage swings from rooftop solar and battery systems by pre-setting how much reactive power each device should supply relative to its active power output. The researchers found an exact mathematical solution for these settings that can perfectly cancel out voltage disruptions when conditions stay within normal bounds, and they mapped out which solar and battery sizes make this solution actually work.

As neighborhoods add more solar panels and batteries, voltage instability threatens equipment damage and blackouts. This approach lets operators stabilize the grid automatically without real-time communication or constant adjustments—crucial for utilities managing thousands of unpredictable devices. The researchers also quantify how much protection is lost if regulations force operators to use weaker control settings, helping regulators decide whether current standards are safe enough.

Explainable Transformer Models for Clinical Prediction Tasks on Structured Electronic Health Records

Making AI predictions from medical records transparent enough for doctors to trust

Researchers built an AI model trained on 75 million patient records that predicts medical outcomes from electronic health records while explaining which specific test results and clinical events drove each prediction. The model performs as well as or better than existing systems on standard medical prediction tasks, and its explanations align with known clinical risk factors—meaning doctors can see why the AI made each recommendation.

Hospitals increasingly rely on AI to flag high-risk patients, but if doctors can't understand the reasoning, they either ignore the alerts or follow them blindly, both dangerous. This model closes that gap by showing which lab values and medical events mattered most for each prediction, letting clinicians verify the logic before acting. The approach works across different medical conditions and prediction types, so it could be deployed broadly across healthcare systems.

Loss-Resilient Semantic Communication over Packet-Loss Networks at Extreme-Low Bandwidth

Sending pictures over terrible internet without losing quality when data packets vanish

Researchers created a new system called ResiGLC that sends images and video over extremely slow internet connections while surviving packet loss—the random data drops that plague poor-quality networks. By borrowing prediction techniques from language models, the system can guess what missing data should be and recover images with better quality than current methods, even when 10–30% of packets disappear in transit.

Extreme-low bandwidth communication matters for disaster zones, remote areas, and satellite links where bandwidth is scarce and connections are unreliable. Current compression systems fail catastrophically when packets drop—a single lost chunk can ruin an entire image. This work means emergency responders, rural clinics, and field researchers can transmit usable images and video over connections that would otherwise be too broken to work with, without needing to compress the data even further and lose critical details.

Multivariate Spatio-Temporal Regression with Penalized Model Selection and an Empirical Application

A statistical toolkit for untangling spatial, temporal, and cross-variable relationships in data

Researchers developed a flexible statistical framework that can detect and model how multiple outcomes influence each other across space and time — for instance, how pollution levels in neighboring cities and past months predict current conditions. Tests on real data from 198 Japanese municipalities showed the method successfully identifies which types of spatial and temporal connections matter most, removing nearly all unexplained spatial patterns from the results.

Policy makers often need to predict outcomes like employment, pollution, or disease spread across regions and years, but existing methods force them to choose between oversimplified models or treating each location and outcome separately. This framework lets analysts automatically discover which connections actually exist in their data, making forecasts more accurate and revealing hidden regional interdependencies that single-location models would miss.

Inject, Align, Recover: Staged Post-Training for Retrieval-Free Document Knowledge Internalization

Teaching AI to remember documents without looking them up during answers

Large language models typically need to retrieve source documents when answering questions about them. Researchers developed a three-stage training method called IAR that lets models internalize document knowledge permanently, so they can answer questions accurately without retrieval — improving domain-specific accuracy by 3.6 percentage points while maintaining general reasoning skills across multiple AI models.

Most AI systems today rely on retrieving relevant documents in real time, which is slow and fails when documents aren't available or are missed by search. A model that permanently learns document content could answer faster and more reliably in settings like customer support, medical records, or legal document systems — where the same documents are queried repeatedly and immediate retrieval isn't always possible.

Auditing Cross-Lingual Fairness in Language Model Watermarking

Why watermarks on AI text work differently across world languages

Watermarking schemes that hide invisible marks in AI-generated text work unevenly across languages, with gaps driven by fundamental linguistic differences rather than random variation. Testing six watermarking methods across eleven languages revealed that performance gaps cluster by language family—not individual languages—suggesting the problem is baked into how these schemes handle different grammatical structures and writing systems.

As AI systems deploy globally, watermarks are a key tool for detecting machine-generated content in moderation and authenticity verification. If watermarks fail silently for speakers of certain languages, some users get strong detection while others face unreliable protection. The framework here provides a concrete way to audit these gaps before deployment, rather than discovering them after problems emerge at scale.

Systemic Risk in Financial Networks Revisited: Debt Dilution as a Backdoor Bail-in

How banks' debt can absorb losses automatically without formal bailouts

Banks can use certain types of debt to automatically shift losses to creditors when financial shocks hit, without needing government intervention or court-ordered restructuring. The researchers found that this "backdoor bail-in" actually makes interconnected banking networks more stable, not less—reversing conventional wisdom that links indebtedness to fragility.

Banks currently rely on government rescues or lengthy legal processes to manage crises, both costly and slow. If interbank debt is structured correctly, losses get absorbed automatically during stress, potentially limiting contagion without taxpayer bailouts. The researchers also explain why real banking networks adopt core-periphery structures: they're harder to destabilize, even if they don't perfectly optimize risk-sharing.

ConceptGuard: Benchmarking Context-Sensitive Unlearning in Large Language Models

Teaching AI to forget bad uses of ideas while keeping good ones

Current methods for removing harmful knowledge from AI systems are too crude—they treat facts as isolated pieces rather than concepts that can be used in multiple ways. Researchers created ConceptGuard, a new benchmark that tests whether AI can eliminate dangerous applications of a concept (like using chemistry for weapons) while preserving safe ones (like using it for medicine), and found that existing unlearning techniques fail this more realistic test.

As AI systems become more powerful, the ability to selectively remove harmful knowledge matters for safety and deployment. Today's unlearning methods can't reliably distinguish between harmful and helpful uses of the same concept, meaning a system might either keep dangerous capabilities intact or strip away genuinely useful knowledge. ConceptGuard provides a concrete way to test whether new safety techniques actually work in real-world scenarios where knowledge has multiple valid and invalid applications.