jeff is a self-hosted drop-in replacement for TypeSafe's jev System One API, powered by the 400M-parameter GLiFormer-large-v1 model. It serves `choice`, `score`, and `noul` classification questions over a compatible wire format, so existing applications using the official `typesafe-sdk` can point at it by changing a single base URL. Deployment targets include GPU (L4, A10G) and CPU (ONNX Runtime with int8 quantization) via Modal, with ~50 req/s per container throughput on L4. Benchmarks on 1,600 labeled items show it costs roughly a quarter of jev per million requests but trails significantly on reasoning-heavy tasks like irony and reading comprehension.
- `JEFF_ISOLATE=nouls` (default) gives separate encoder passes per noul question to reduce cross-question interference; choice and score questions share a pass unless set to `all`
- Default temperature of 3.2 calibrates noul probabilities; setting it to 1 makes `score` output match the weighted average of displayed probabilities
- A smaller `gliformer-base-v1` variant with `JEFF_NOUL_MODE=single` is recommended for faster local iteration
- `JEFF_ISOLATE=nouls` (default) gives separate encoder passes per noul question to reduce cross-question interference; choice and score questions share a pass unless set to `all`
- Default temperature of 3.2 calibrates noul probabilities; setting it to 1 makes `score` output match the weighted average of displayed probabilities
- A smaller `gliformer-base-v1` variant with `JEFF_NOUL_MODE=single` is recommended for faster local iteration
Carolina Bento writes about Linear Discriminant Analysis (LDA), a supervised learning technique used for dimensionality reduction and pattern recognition. The article explains how LDA maximizes class separability by maximizing the ratio of between-class to within-class variance, making it particularly useful for simplifying high-dimensional datasets while preserving core characteristics. Through a real estate dataset example, the author demonstrates how to implement LDA using ScikitLearn to visualize property type clusters and identify key features that distinguish different types of properties.
- LDA is a supervised method, unlike Principal Component Analysis (PCA), which is unsupervised.
- The maximum number of Linear Discriminants that can be calculated is $K-1$, where $K$ is the number of classes.
- Key assumptions for LDA include linear separability of data and following a Gaussian distribution.
- It helps reduce overfitting by removing redundant features and minimizing noise in high-dimensional spaces.
- LDA is a supervised method, unlike Principal Component Analysis (PCA), which is unsupervised.
- The maximum number of Linear Discriminants that can be calculated is $K-1$, where $K$ is the number of classes.
- Key assumptions for LDA include linear separability of data and following a Gaussian distribution.
- It helps reduce overfitting by removing redundant features and minimizing noise in high-dimensional spaces.
Benjamin Nweke writes that traditional fraud detection relies on the assumption of a human actor, where deviations from established behavioral patterns serve as primary signals. While explainability tools like SHAP can effectively detail why specific transaction features (like amount or timing) trigger a risk score, they are insufficient for addressing "machine-to-machine mayhem" caused by autonomous agents. Because these agents lack human biological constraints and consistent life patterns, feature attribution on transactions fails to capture the underlying intent or decision-making trajectory of an agent that may be operating outside its delegated scope.
- Agentic AI fraud is characterized as a shift toward "machine-to-machine mayhem" where bots mimic legitimate shopping agents.
- Current explainability methods like SHAP focus on transaction features rather than the actor's underlying decision path or tool usage.
- 60% of industry professionals expect AI-mediated banking to diminish the effectiveness of traditional fraud defenses.
- Proposed regulatory responses include NIST's Agent Standards Initiative and Senator Mark Warner's proposed AI AGENT Act for establishing accountability through registries.
- Agentic AI fraud is characterized as a shift toward "machine-to-machine mayhem" where bots mimic legitimate shopping agents.
- Current explainability methods like SHAP focus on transaction features rather than the actor's underlying decision path or tool usage.
- 60% of industry professionals expect AI-mediated banking to diminish the effectiveness of traditional fraud defenses.
- Proposed regulatory responses include NIST's Agent Standards Initiative and Senator Mark Warner's proposed AI AGENT Act for establishing accountability through registries.
Cheng Xue writes that cognitive capacity limitations and reduced performance under uncertain conditions stem from feature interference, where irrelevant features become strongly represented or entangled with relevant ones in neural representations. By combining monkey electrophysiology, human psychophysics, and artificial neural network modeling, researchers demonstrated that both humans and monkeys exhibit decreased perceptual accuracy when task rules are uncertain compared to optimized networks. This phenomenon is driven by the induction of stronger-than-normal representations for irrelevant features during periods of uncertainty or after nonrewarded trials.
- The study found that unrewarded trials lead to higher feature interference in neural populations
- Monkey choice models trained on behavioral data replicated suboptimal performance seen in live subjects, whereas "correct" networks did not
- Feature axes become less orthogonal (more entangled) in the monkey choice network following nonrewards
- Task certainty can be continuously measured by tracking task output activity in recurrent neural networks
- The study found that unrewarded trials lead to higher feature interference in neural populations
- Monkey choice models trained on behavioral data replicated suboptimal performance seen in live subjects, whereas "correct" networks did not
- Feature axes become less orthogonal (more entangled) in the monkey choice network following nonrewards
- Task certainty can be continuously measured by tracking task output activity in recurrent neural networks
This repository provides an open-source face recognition software development kit (SDK) for Windows and Linux systems, developed by Faceplugin. It uaes deep learning models to offer on-premise processing of facial data, ensuring privacy as no information leaves the user's device. The toolkit supports various functions including face detection, landmark detection, feature embedding generation, and similarity comparison via Python APIs.
- Supports JPG, PNG, BMP, and TIFF image formats
- Compatible with both CPU and GPU acceleration
- Requires Python 3.9 or higher and Anaconda is recommended for setup
- Includes capabilities for bounding box extraction and facial landmark detection
- Supports JPG, PNG, BMP, and TIFF image formats
- Compatible with both CPU and GPU acceleration
- Requires Python 3.9 or higher and Anaconda is recommended for setup
- Includes capabilities for bounding box extraction and facial landmark detection
Hugging Face Unveils Microduck: A $399 Open-Source 25 cm Biped You Train with Reinforcement LearningMichal Sutter writes about Pollen Robotics, a Bordeaux-based team at Hugging Face, which has opened pre-orders for Microduck, a 25 cm bipedal robot priced at $399. Unlike most robotics launches that rely on demo videos, Microduck ships with its full training loop — every movement (walking, sitting, kicking, roller-skating, self-recovery) is a neural policy trained in a physics simulator and exported to hardware. The robot carries 15 motors, a camera, LiDAR, two IMUs, and a Rockchip RK3566, with policies trained via PPO in MuJoCo Warp in roughly one to two hours on a CUDA GPU.
- Sim-to-real hinges on a BAM actuator model (voltage control law, back-EMF, Coulomb/Stribeck/load-dependent friction) plus randomization of battery voltage, command delay, and ±1° backlash per joint
- Every policy shares a 61-dimensional actor observation (48 proprioception + twist, head pose, body pose commands), enabling hot-swap between walk, recover, and trick policies mid-run
- Software is Apache-2.0, but mechanical and electronic design files are not open
- The robot generates a unique audio identity on first wake that persists permanently; it does not speak in a linguistic sense
- Pre-orders opened August 27, 2026, with deliveries targeted before Christmas
- Sim-to-real hinges on a BAM actuator model (voltage control law, back-EMF, Coulomb/Stribeck/load-dependent friction) plus randomization of battery voltage, command delay, and ±1° backlash per joint
- Every policy shares a 61-dimensional actor observation (48 proprioception + twist, head pose, body pose commands), enabling hot-swap between walk, recover, and trick policies mid-run
- Software is Apache-2.0, but mechanical and electronic design files are not open
- The robot generates a unique audio identity on first wake that persists permanently; it does not speak in a linguistic sense
- Pre-orders opened August 27, 2026, with deliveries targeted before Christmas
Iván Palomares Carrascosa writes about methods for interpreting the dense numerical vector representations, or embeddings, generated by large language models (LLMs). By using a combination of probing classifiers like logistic regression, UMAP dimensionality reduction for visualization, and SHAP values to identify influential latent dimensions, one can analyze the quality and semantic structure captured within LLM-generated embedding spaces.
- Probing classifiers help determine if embeddings are rich enough to distinguish between classes by testing them with simpler models.
- UMAP is used to project high-dimensional embeddings into 2D space for visual inspection of natural groupings.
- SHAP values can pinpoint which specific dimensions in an embedding most significantly influence a classifier's decisions.
- The article demonstrates using Scikit-LLM alongside local Ollama models to generate embeddings cost-effectively.
- Probing classifiers help determine if embeddings are rich enough to distinguish between classes by testing them with simpler models.
- UMAP is used to project high-dimensional embeddings into 2D space for visual inspection of natural groupings.
- SHAP values can pinpoint which specific dimensions in an embedding most significantly influence a classifier's decisions.
- The article demonstrates using Scikit-LLM alongside local Ollama models to generate embeddings cost-effectively.
Chris Patrick writes that SLAC researchers have built a neural-network method for compressing large scientific datasets while preserving fine details that conventional compression erases. The approach uses wavelet analysis to separate data features by scale, then encodes each scale separately through a neural network, enabling 10- to 100-fold file size reductions and selective decompression of only the regions a researcher needs.
- Published in Nature Machine Intelligence (August 24, 2026)
- Motivated by upcoming LCLS upgrades that will generate nearly one terabyte of data per second
- Tested successfully on X-ray diffraction data, solar magnetic field measurements, and photographs
- Neural networks trained on Perlmutter at NERSC (Lawrence Berkeley National Lab)
- Co-developers include researchers from UC Davis and Carnegie Mellon University
- Published in Nature Machine Intelligence (August 24, 2026)
- Motivated by upcoming LCLS upgrades that will generate nearly one terabyte of data per second
- Tested successfully on X-ray diffraction data, solar magnetic field measurements, and photographs
- Neural networks trained on Perlmutter at NERSC (Lawrence Berkeley National Lab)
- Co-developers include researchers from UC Davis and Carnegie Mellon University
El Assadi et al. compare ten LLMs (six families) and 26 embedding models (118M - 14B parameters) on 37 tasks, considering cost. In aggregate, the two paradigms are effectively tied (best LLM scores 77.6 versus best embedding model 77.2), yet their strengths diverge by task: LLMs lead on reasoning-heavy retrieval while embedding models lead on classification, and the two match on clustering, STS, and pair classification.
LLMs are significantly more expensive (up to 1,431x) and slower (2.5-736x) than embedding models for certain tasks. The authors suggest using embedding models for similarity, classification, and clustering, and LLMs for reasoning in retrieval.
Reasoning tokens are 28-81% of LLM inference cost; lower budgets maintain or boost retrieval quality for most tested models.
- Only Gemini 3.1 Pro breaks into the Pareto frontier alongside the leading embedding models.
- Accepted to COLM 2026; code, datasets, and results are publicly released on GitHub.
LLMs are significantly more expensive (up to 1,431x) and slower (2.5-736x) than embedding models for certain tasks. The authors suggest using embedding models for similarity, classification, and clustering, and LLMs for reasoning in retrieval.
Reasoning tokens are 28-81% of LLM inference cost; lower budgets maintain or boost retrieval quality for most tested models.
- Only Gemini 3.1 Pro breaks into the Pareto frontier alongside the leading embedding models.
- Accepted to COLM 2026; code, datasets, and results are publicly released on GitHub.
Alibaba has open-sourced Qwen-UI-Agent, a GUI agent foundation model that operates across mobile, desktop, web, and deep-search environments on real hardware rather than relying on simulation. It achieves top benchmark results: 82.1% on MobileWorld, 79.5% on OSWorld-Verified, and first on WebArena. It also introduces MobileWorld-Real, a 400+ task benchmark on 100+ phones and 150+ apps, with a 92.2% success rate.
- Supports command-line execution alongside standard GUI operations and batches multiple actions into a single decision step to shorten trajectories.
- Built-in safety layer refuses illegal or high-risk requests outright and pauses at sensitive operations (payments, data deletion, privacy grants) for explicit user confirmation.
- Trained via online reinforcement learning on trajectories exceeding 100 steps, paired with adaptive curriculum learning to progressively tackle longer tasks.
- Supports command-line execution alongside standard GUI operations and batches multiple actions into a single decision step to shorten trajectories.
- Built-in safety layer refuses illegal or high-risk requests outright and pauses at sensitive operations (payments, data deletion, privacy grants) for explicit user confirmation.
- Trained via online reinforcement learning on trajectories exceeding 100 steps, paired with adaptive curriculum learning to progressively tackle longer tasks.
