AI BriefingSamsung Research news
The latest AI and developer news from Samsung Research, with short summaries.
Visit website →Samsung Research Proposes Systematic Probabilistic Shaping for 6G Modems
Samsung Research proposes Systematic Probabilistic Shaping (S-PS), which shapes the second-most-significant bit of QAM symbols to improve spectral efficiency while reusing existing 5G NR LDPC coding and modulation hardware.
Samsung Research Demonstrates ISAC on Commercial 5G vRAN Using Uplink CSI for Human Sensing
The system reuses existing uplink reference signals to distinguish four presence states without additional hardware, reducing data volume by over 10x.
Samsung Introduces CG-TTRL for Context-Guided On-Device LLM Adaptation
The framework improves reasoning accuracy in 1.5B parameter models by integrating In-Context Learning into Test-Time Reinforcement Learning.
Samsung Research Details Ambient IoT Design for Battery-Free 6G Connectivity
The article defines Ambient IoT as a RAT-independent interface for devices consuming 1–100 microwatts, enabling network-managed inventory without dedicated readers.
Samsung R&D Releases langgraph-goat for Explainable Multi-Agent AI Systems
The open-source library implements the Graph of Agentic Thoughts framework to capture decision reasoning and rejected alternatives in LangGraph applications.
3GPP Release 20 Standardizes Voice over NB-IoT NTN for Satellite Connectivity
The standard targets geostationary satellites with a new ultra-low bitrate codec operating at 1.0 and 1.4 kbps.
Samsung Electronics Unveils 'MusicJudge', an AI Song Evaluation Framework Reflecting Musical Structure
Samsung Electronics has released an AI framework that analyzes lyrics and musical structure together to evaluate songs at a human-expert level.
Samsung Electronics Releases AnySimLite, a 700KB Ultra-Lightweight Voice Classification Encoder
Samsung Electronics has released AnySimLite, an ultra-lightweight voice classification encoder capable of sub-30ms inference on smartphones with a size of 700KB.
Samsung Electronics Proposes Strategy to Ensure Mobile Service Continuity Outside Terrestrial Coverage via 6G NTN
Samsung Electronics defines 6G NTN not as a replacement for terrestrial networks but as a service continuity layer, presenting a technical foundation to ensure essential mobile services remain uninterrupted outside terrestrial coverage.
Simplified, Efficient, and Intelligent User Plane Design for 6G
The 6G user plane will maximize latency and energy efficiency by simplifying the existing hierarchical structure and introducing intelligent processing.
AI-Based Optimization for User and Service-Level Perceived Quality
Samsung Research proposed a technology that uses AI to predict and proactively control individual user quality degradation in 5G-Advanced and 6G networks.
Maximizing Uplink Potential: AI-Based Channel Selectivity Matching Using Compressed Channel State Information
Samsung proposed a method to improve uplink efficiency using compressed CSI and AI-based precoding.
Neural Network Bit Recovery for FDMA-Based Wide-Area Ambient-IoT Networks
Samsung improved bit recovery efficiency for Ambient-IoT tags using FDMA and Hopfield Neural Networks.
LittleBit-2: Maximizing Spectral Energy Gain in Sub-1-Bit LLMs via Latent Geometric Alignment
LittleBit-2 improves the quantization performance of sub-1-bit LLMs through latent geometric alignment.
NanoQuant: Efficient Sub-1-Bit Quantization for Large Language Models
NanoQuant compresses a 70B LLM to below 1 bit using a single H100 GPU.
RaBiT: A Residual-Aware Binarization Training Method for Accurate and Efficient LLMs
RaBiT is proposed as a binarization training technique that leverages residuals to improve accuracy while maximizing the efficiency of LLMs.
Phase-Error-Robust Practical Channel Estimation Using Special Orthogonal Group Equivariant AI/ML
Samsung R&D has proposed SOGeCE, a low-complexity AI channel estimation technique that leverages Special Orthogonal Group (SOG) equivariant structure to simultaneously handle phase errors and additive noise.
PickCross-Biosignal Pretraining for Wearable Health AI: Reflecting the Underlying Biosignal Structure
Samsung Research America proposed xMAE, a cross-modal pretraining framework that uses ECG as a training aid signal to enhance PPG representational power.
Large-Scale AI CSI Compression for 6G FR3
Samsung Research proposed a two-stage AI CSI feedback training method for extreme compression in 256-port X-MIMO environments.
K-Merge: Online Continual Merging of Adapters for On-Device LLMs
We propose the K-Merge framework, which continually expands an LLM's capabilities without data even in storage-constrained on-device environments.