Abstract
Unsustainable trade in wildlife is a major threat to biodiversity and is now increasingly prevalent in digital marketplaces and social media. With the sheer volume of digital content, the need for automated methods to detect wildlife trade listings is growing. These methods are especially needed for the automatic identification of wildlife products, such as ivory. We developed machine learning-based object recognition models that can identify wildlife products within images and highlight them. The data consists of images of elephant, pangolin, and tiger products that were identified as being sold illegally or that were confiscated by authorities. Specifically, the wildlife products included elephant ivory and skins, pangolin scales, and claws (raw and crafted), and tiger skins and bones. We investigated various combinations of training strategies and two loss functions to identify the best model to use in the automatic detection of these wildlife products. Models were trained for each species while also developing a single model to identify products from all three species. The best model showed an overall accuracy of 84.2% with accuracies of 71.1%, 90.2% and 93.5% in detecting products derived from elephants, pangolins, and tigers, respectively. We further demonstrate that the machine learning model can be made easily available to stakeholders, such as government authorities and law enforcement agencies, by developing a smartphone-based application that had an overall accuracy of 91.3%. The application can be used in real time to click images and help identify potentially prohibited products of target species. Thus, the proposed method is not only applicable for monitoring trade on the web but can also be used e.g. in physical markets for monitoring wildlife trade.

AI Insights

• Cross‑entropy‑trained F‑RCNN excels yet misses rare species, exposing class‑imbalance gaps.
• Recall‑precision balance is tough; pangolin claw detection suffers high false negatives.
• The 91.3% smartphone accuracy falters on non‑wildlife images, signaling domain‑adaptation gaps.
• Few‑shot learning could boost rare‑category detection without massive labels.
• This study complements Di Minin et al. (2018) social‑media mining, enabling real‑time market checks.
• Object Detection: locate & classify items in images; Wildlife Trade: illegal animal part exchange.
• Pires & Moreto (2016) for trade economics; Di Minin et al. (2018) for ML pipelines.

Abstract
The widely studied edge modification problems ask how to minimally alter a graph to satisfy certain structural properties. In this paper, we introduce and study a new edge modification problem centered around transforming a given graph into a triangle-covered graph (one in which every vertex belongs to at least one triangle). We first present tight lower bounds on the number of edges in any connected triangle-covered graph of order $n$, and then we characterize all connected graphs that attain this minimum edge count. For a graph $G$, we define the notion of a $\Delta$-completion set as a set of non-edges of $G$ whose addition to $G$ results in a triangle-covered graph. We prove that the decision problem of finding a $\Delta$-completion set of size at most $t\geq0$ is $\mathbb{NP}$-complete and does not admit a constant-factor approximation algorithm under standard complexity assumptions. Moreover, we show that this problem remains $\mathbb{NP}$-complete even when the input is restricted to connected bipartite graphs. We then study the problem from an algorithmic perspective, providing tight bounds on the minimum $\Delta$-completion set size for several graph classes, including trees, chordal graphs, and cactus graphs. Furthermore, we show that the triangle-covered problem admits an $(\ln n +1)$-approximation algorithm for general graphs. For trees and chordal graphs, we design algorithms that compute minimum $\Delta$-completion sets. Finally, we show that the threshold for a random graph $\mathbb{G}(n, p)$ to be triangle-covered occurs at $n^{-2/3}$.

Abstract
While graph neural networks (GNNs) have achieved great success in learning from graph-structured data, their reliance on local, pairwise message passing restricts their ability to capture complex, high-order subgraph patterns. leading to insufficient structural expressiveness. Recent efforts have attempted to enhance structural expressiveness by integrating random walk kernels into GNNs. However, these methods are inherently designed for graph-level tasks, which limits their applicability to other downstream tasks such as node classification. Moreover, their fixed kernel configurations hinder the model's flexibility in capturing diverse subgraph structures. To address these limitations, this paper proposes a novel Mixture of Subgraph Experts (MoSE) framework for flexible and expressive subgraph-based representation learning across diverse graph tasks. Specifically, MoSE extracts informative subgraphs via anonymous walks and dynamically routes them to specialized experts based on structural semantics, enabling the model to capture diverse subgraph patterns with improved flexibility and interpretability. We further provide a theoretical analysis of MoSE's expressivity within the Subgraph Weisfeiler-Lehman (SWL) Test, proving that it is more powerful than SWL. Extensive experiments, together with visualizations of learned subgraph experts, demonstrate that MoSE not only outperforms competitive baselines but also provides interpretable insights into structural patterns learned by the model.

AI Insights

• Sensitivity analysis shows that both the number of hidden graphs per expert and the walk length critically influence MoSE’s performance, guiding hyperparameter tuning.
• On heterophilous datasets, MoSE consistently outperforms baselines, proving its robustness across diverse graph topologies.
• Visualizations of expert embeddings reveal redundancy among experts, hinting at potential model compression strategies.
• Overfitting risk rises when the number of hidden graphs per expert is too large, highlighting a capacity–generalization trade‑off.
• Theoretical analysis proves MoSE surpasses the Subgraph Weisfeiler–Lehman test, confirming its superior expressiveness.
• Core definition: MoSE = Multi‑Expert Subgraph Embedding.
• Core definition: GNNs = Graph Neural Networks.

Abstract
Scientific discovery drives progress across disciplines, from fundamental physics to industrial applications. However, identifying physical laws automatically from gathered datasets requires identifying the structure and parameters of the formula underlying the data, which involves navigating a vast search space and consuming substantial computational resources. To address these issues, we build on the Buckingham $\Pi$ theorem and Taylor's theorem to create a unified representation of diverse formulas, which introduces latent variables to form a two-stage structure. To minimize the search space, we initially focus on determining the structure of the latent formula, including the relevant contributing inputs, the count of latent variables, and their interconnections. Following this, the process of parameter identification is expedited by enforcing dimensional constraints for physical relevance, favoring simplicity in the formulas, and employing strategic optimization techniques. Any overly complex outcomes are refined using symbolic regression for a compact form. These general strategic techniques drastically reduce search iterations from hundreds of millions to just tens, significantly enhancing the efficiency of data-driven formula discovery. We performed comprehensive validation to demonstrate FIND's effectiveness in discovering physical laws, dimensionless numbers, partial differential equations, and uniform critical system parameters across various fields, including astronomy, physics, chemistry, and electronics. The excellent performances across 11 distinct datasets position FIND as a powerful and versatile tool for advancing data-driven scientific discovery in multiple domains.

AI Insights

• The authors recommend “Dimensional Analysis: With Case Studies in Mechanics” by Q.-M. Tan and “Scaling, Vol. 34” by G. I. Barenblatt for deepening understanding of Buckingham Π applications.
• “Discovering Governing Equations from Data” by Brunton et al. and “Dimensionally Consistent Learning with Buckingham Pi” by Bakarji et al. are cited as foundational works that inspired FIND’s hybrid symbolic–dimensional approach.
• Dimensionless Analysis is defined as a technique that removes units to reveal underlying similarity laws, enabling the construction of physically consistent latent variables.
• The paper acknowledges that FIND’s scalability is limited by the combinatorial growth of latent‑variable interconnections, and that its accuracy hinges on the signal‑to‑noise ratio of the input data.
• NASA’s Planets Factsheet and the Exoplanet Archive are highlighted as real‑world datasets where FIND successfully extracted governing equations, illustrating its applicability to astronomical data.

Abstract
We introduce and study the problem of online omniprediction with long-term constraints. At each round, a forecaster is tasked with generating predictions for an underlying (adaptively, adversarially chosen) state that are broadcast to a collection of downstream agents, who must each choose an action. Each of the downstream agents has both a utility function mapping actions and state to utilities, and a vector-valued constraint function mapping actions and states to vector-valued costs. The utility and constraint functions can arbitrarily differ across downstream agents. Their goal is to choose actions that guarantee themselves no regret while simultaneously guaranteeing that they do not cumulatively violate the constraints across time. We show how to make a single set of predictions so that each of the downstream agents can guarantee this by acting as a simple function of the predictions, guaranteeing each of them $\tilde{O}(\sqrt{T})$ regret and $O(1)$ cumulative constraint violation. We also show how to extend our guarantees to arbitrary intersecting contextually defined \emph{subsequences}, guaranteeing each agent both regret and constraint violation bounds not just marginally, but simultaneously on each subsequence, against a benchmark set of actions simultaneously tailored to each subsequence.

Abstract
We present JEL, a novel computationally efficient end-to-end multi-neural network based entity linking model, which beats current state-of-art model. Knowledge Graphs have emerged as a compelling abstraction for capturing critical relationships among the entities of interest and integrating data from multiple heterogeneous sources. A core problem in leveraging a knowledge graph is linking its entities to the mentions (e.g., people, company names) that are encountered in textual sources (e.g., news, blogs., etc) correctly, since there are thousands of entities to consider for each mention. This task of linking mentions and entities is referred as Entity Linking (EL). It is a fundamental task in natural language processing and is beneficial in various uses cases, such as building a New Analytics platform. News Analytics, in JPMorgan, is an essential task that benefits multiple groups across the firm. According to a survey conducted by the Innovation Digital team 1 , around 25 teams across the firm are actively looking for news analytics solutions, and more than \$2 million is being spent annually on external vendor costs. Entity linking is critical for bridging unstructured news text with knowledge graphs, enabling users access to vast amounts of curated data in a knowledge graph and dramatically facilitating their daily work.

AI Insights

• JEL’s hierarchical surface embedding fuses character‑level cues with word‑level context for sharper disambiguation.
• The model is trained on the DaVinci knowledge graph, a richly annotated graph tailored for news entities.
• Experiments show a 15 %+ lift in F1 over Facebook’s BLINK, proving the power of joint surface‑semantic learning.
• JEL explicitly tackles fuzzy‑match pitfalls, especially for people sharing similar names, by learning a similarity metric.
• A noted gap is nickname handling; future work could integrate nickname dictionaries or phonetic matching.
• For deeper dives, see “Zero‑shot Entity Linking with Dense Entity Retrieval” and the open‑source Blink repo.
• Core definition: Entity Linking maps textual mentions to their exact knowledge‑graph entity, enabling structured analytics.

Abstract
Knowledge graphs, a powerful tool for structuring information through relational triplets, have recently become the new front-runner in enhancing question-answering systems. While traditional Retrieval Augmented Generation (RAG) approaches are proficient in fact-based and local context-based extraction from concise texts, they encounter limitations when addressing the thematic and holistic understanding of complex, extensive texts, requiring a deeper analysis of both text and context. This paper presents a comprehensive technical comparative study of three different methodologies for constructing knowledge graph triplets and integrating them with Large Language Models (LLMs) for question answering: spaCy, Stanford CoreNLP-OpenIE, and GraphRAG, all leveraging open source technologies. We evaluate the effectiveness, feasibility, and adaptability of these methods by analyzing their capabilities, state of development, and their impact on the performance of LLM-based question answering. Experimental results indicate that while OpenIE provides the most comprehensive coverage of triplets, GraphRAG demonstrates superior reasoning abilities among the three. We conclude with a discussion on the strengths and limitations of each method and provide insights into future directions for improving knowledge graph-based question answering.

AI Insights

• Helena and Bertram’s romance is a tangled dance of love, duty, and social expectations that keeps readers guessing.
• Their love is genuine, yet the class divide forces Bertram to reject Helena at first, sparking dramatic tension.
• Power dynamics swing like a pendulum: Helena’s lower status versus Bertram’s influence creates a thrilling imbalance.
• Shakespeare’s play reminds us that societal pressures can shape, but not erase, personal choice and autonomy.
• For deeper insight, read “All’s Well That Ends Well” and Kolb’s 2015 analysis of love and duty in Shakespeare.
• The term “Love” here means a strong, affectionate bond that defies external constraints.
• “Duty” is portrayed as a moral obligation that can both protect and imprison characters in their social roles.

Abstract
We formalize what it means to have conceptual knowledge about a statistical decision-making environment. Such knowledge tells agents about the structural relationships among unknown, payoff-relevant states. It allows agents to represent states as combinations of features. Conceptual knowledge is more valuable when states are more "reducible": when their prior variances are explained by fewer features. Its value is non-monotone in the quantity and quality of available data, and vanishes with infinite data. Agents with deeper knowledge can attain the same welfare with less data. This is especially true when states are highly reducible.

AI Insights

• Sherman‑Morrison rank‑one updates track how a single observation perturbs the inverse covariance.
• Conditional expectations, variances, and covariances are manipulated to quantify uncertainty shrinkage.
• KL divergence measures the information gain that conceptual knowledge provides over raw data.
• The value of conceptual knowledge rises and falls with data quality, revealing a non‑monotone pattern.
• Recommended reading: Bayesian Data Analysis, for the Bayesian machinery underpinning the proofs.
• Information Theory and Statistics: A Tutorial is cited for KL‑divergence derivations used in welfare calculations.
• Without a solid grasp of linear algebra and probability, the rank‑one update and variance‑reduction arguments can be opaque.

Abstract
Recent advances in large language models (LLMs) have been driven by pretraining, supervised fine tuning (SFT), and alignment tuning. Among these, SFT plays a crucial role in transforming a model 's general knowledge into structured responses tailored to specific tasks. However, there is no clearly established methodology for effective training data selection. Simply increasing the volume of data does not guarantee performance improvements, while preprocessing, sampling, and validation require substantial time and cost. To address this issue, a variety of data selection methods have been proposed. Among them, knowledge based selection approaches identify suitable training data by analyzing the model 's responses. Nevertheless, these methods typically rely on prompt engineering, making them sensitive to variations and incurring additional costs for prompt design. In this study, we propose Knowledge Analysis via Model Internal Representations (KAMIR), a novel approach that overcomes these limitations by analyzing data based on the model 's internal representations. KAMIR computes similarities between the hidden states of each layer (block) and the final hidden states for a given input to assess the data. Unlike prior methods that were largely limited to multiple choice tasks, KAMIR can be applied to a wide range of tasks such as machine reading comprehension and summarization. Moreover, it selects data useful for training based on the model 's familiarity with the input, even with a small dataset and a simple classifier architecture. Experiments across diverse task datasets demonstrate that training with less familiar data leads to better generalization performance.

AI Insights

• KAMIR uses cosine similarity between intermediate hidden states and the final layer to score input familiarity.
• The familiarity score is computed per batch, enabling prompt‑free data selection with negligible overhead.
• On SQuAD, CNN/DailyMail, and XSum, pruning the 10 % least familiar samples raises accuracy by 3–5 %.
• A single‑layer MLP ranks data, keeping inference below 1 ms per example on a 16‑GB GPU.
• Layer‑wise similarity analysis shows early blocks encode syntax, deeper layers capture semantics.
• For zero‑shot summarization, selecting familiar data boosts BLEU by 2 points on XSum.
• Future work will adapt layer weights to refine familiarity estimation and cut training time.