On-chain behavioral signatures classify centralized exchange addresses
LightGBM outperforms logistic regression reaching PR-AUC 0.9594 on 3 million Ethereum addresses.
Stories per day · 30 days
2 RSS
Relatedattention1blockchain1cascading1data-discovery1decision-models1dspm1efficiency1email-agents1ethereum1gemini1knowledge-conflict1llm-classifiers1llm-evaluation1logistic-regression1microsoft-purview1multilingual1proof-of-reserves1query-cost1
23 stories, newest first · one card per story
Stories tagged #classification
Buyer's guide compares 12 data discovery tools for 2026
The guide weighs detection honesty, identity correlation, estate coverage, remediation hooks, and scan cost.
1
Decision models type RAG source conflicts cheaper than Gemini
Eleven decision models hit 50.0–67.0% on CONFLICTS, and 10 matched or beat Gemini 2.5 Flash at 4–39x lower billed cost.
Foyer study: logistic regression beats LLMs on inbox routing
On about 200 labels, logistic regression led every field, while the most accurate system ignored 18% of emails needing action.
LLM classifier circuits raise accuracy at a fixed query cost
On seven tasks spanning safety, health, and reasoning, the method beats routing and cascading baselines.
MJev-Spatial: System One Spatial Intelligence via Native Attention Matching
MJev-Spatial classifies spatial answers on Qwen3.5-4B without decoding tokens, topping twelve benchmarks and running much faster.
What is Decision 3.0? vLLM Semantic Router’s New Open Decision Models
vLLM Semantic Router released Decision 3.0, five Apache-2.0 multimodal models from 0.8B to 27B.
OpenAI launches Decisions API that reduces complex evaluations to yes, no, or pick one
OpenAI’s Decisions API, in public beta on gpt-6-luna, returns yes/no, category, or scale judgments.
Conformal Prediction Sets Quantify Information Gain: A Theoretical Perspective
Paper shows conformal prediction set-size reduction tracks information gain and relates to Shannon mutual information.
Decision AI Models Explained: TypeSafe Jev vs Fastino GLiDE, GLiNER2.5-Decide and Open-Source Competitors
TypeSafe's Jev and rivals return calibrated choices and scores instead of text, at much lower cost and latency.
13
A Structured State Space Sequence Model for Multi-Class Classification of Malware
Researchers apply a Structured State Space model to classify IoT malware from long-range execution dependencies.
Robust Non-Clairvoyant Scheduling with Classification Models
A confusion-matrix robustness framework yields optimal non-adaptive scheduling when job times are unknown until completion.
Labels Override Definitions in Jev-Style Typed Decision Models
Open typed decision models mostly follow option labels, not definitions, because of prompt rendering.
Asymptotic Properties of Support Vector Machines in High-Dimension, Low-Sample-Size Settings under a Spiked Model
Under a spiked HDLSS model, standard SVM is inconsistent; a spike-corrected SVM needs growing sample size.
Liquid AI Releases d1: A Decision Model That Returns Calibrated Probabilities With Zero Output Tokens
Liquid AI launched d1, a hosted decision model returning calibrated probabilities with zero output tokens.
ProtoSeam: Lifting Classifier Training with Latent Gaussian Mixture Models
ProtoSeam lifts classifier training with class prototypes, gaining up to five points without changing inference.
Fastino Releases GLiNER2.5-Decide: A 340M Open-Weight Decision Model That Runs on CPU
Fastino Labs released GLiNER2.5-Decide, a 340M Apache 2.0 open-weight decision model for agent routing, triage, and guardrails that runs on CPU.
SupersonicLabs/Julia-1 — new model trending #30 on Hugging Face
Supersonic Labs released Julia 1, a 144.3M-parameter model that picks among supplied answers for classification and routing.
3
A Bulletproof Business? Towards Detecting Infrastructure-as-a-Service Offerings on Telegram
Researchers detect cybercrime infrastructure-as-a-service ads across more than one million Telegram messages.
LLM Classification Is Feature Engineering
Argues LLM classifiers should feed a downstream logistic regression, yielding calibration, threshold control, and principled use of structured covariates.
2
Detecting Deceptive Recruitment: A Signal-theoretic Machine Learning Framework for Early Identification of Labour Exploitation
Researchers build a multimodal ML framework detecting deceptive job advertisements used for labour exploitation, achieving ROC-AUC 0.87–0.97 across individual modalities.
Former OpenAI researcher builds an AI model that judges options instead of writing text
TypeSafe AI launches Jev, a judgment-only model built by ex-OpenAI staff that classifies inputs with 70-500 ms latency instead of generating text.
6
Optimal Rates for Agentic Networked Information Aggregation
Researchers close the Kearns–Roth–Ryu gap for agentic networked information aggregation, proving excess error is constant up to depth M^2 then Θ(M^2/D).