ZK
Editor-in-Chief · CPO of AlpinaGPT

Zhemal Khamidun

Head of AI at Alpina Digital, CPO of AlpinaGPT (~30-person engineering & AI team). Founder of Vibe Academy & Hamidun Consulting, author of the AI channel «Готовим ИИшницу» (@jemal_hamidun).

10,654Published articles
50+Verified primary sources
6Language editions (RU/EN/ES/PT/FR/AR)
Researchers Prove Why Adversarial Training Makes Saliency Maps Sparse
Research·arXiv cs.LG

Researchers Prove Why Adversarial Training Makes Saliency Maps Sparse

Oct 9, 2026·2 min
Distilling ECG into PPG Memorized Identity: Fixed-Effects More Than Doubled the State Signal
Medicine

Distilling ECG into PPG Memorized Identity: Fixed-Effects More Than Doubled the State Signal

Oct 9, 2026·3 min
Subliminal learning passed a backdoor through numbers: 23.5% of replies in French
Research

Subliminal learning passed a backdoor through numbers: 23.5% of replies in French

Oct 9, 2026·2 min
PXtal Beats Baseline Models on Six Benchmarks, Four of Them Zero-Shot
Research

PXtal Beats Baseline Models on Six Benchmarks, Four of Them Zero-Shot

Oct 9, 2026·2 min
Nullify Erases Knowledge From LLMs Without Retraining, Using Null-Space Vectors
Research·arXiv cs.LG

Nullify Erases Knowledge From LLMs Without Retraining, Using Null-Space Vectors

Nullify makes an LLM "forget" private data at inference time: the weights stay unchanged, and forgetting quality on TOFU and MUSE matches baseline methods.

Oct 9, 2026·2 min
Multimodal Alzheimer's AI model reaches 0.935 AUROC on 2649 patients
Medicine·arXiv cs.LG

Multimodal Alzheimer's AI model reaches 0.935 AUROC on 2649 patients

A new framework combines MRI, clinical transformers and ODE-GRU. On ADNI, OASIS-2 and MIRIAD, diagnostic AUROC is 0.935, but on the external OASIS-3 cohort it drops…

Oct 9, 2026·2 min
CARing teaches LLMs to predict diagnoses: R@30 hits 46.52% on MIMIC-IV
Research·arXiv cs.LG

CARing teaches LLMs to predict diagnoses: R@30 hits 46.52% on MIMIC-IV

CARing's authors encode diagnoses as Semantic IDs and reward the model for covering distinct diagnoses. In reasoning mode, R@30 is 46.04% on MIMIC-III and 46.52% on…

Oct 9, 2026·2 min
Visible CoT Does Not Universally Improve SQL and pandas Code Generation, New Study Finds
LLM·arXiv cs.LG

Visible CoT Does Not Universally Improve SQL and pandas Code Generation, New Study Finds

An arXiv preprint tested the "think out loud" advice on SQL and pandas with code execution. The authors found no general CoT gain and no benefit from "thinking in…

Oct 9, 2026·2 min
Top themes
01D-SLR Replaces Truncated SVD With Three Decompositions and No TuningOct 9, 2026·2 min02Phase-HDC trains a model with no optimizer history and stores 16–23x less than AdamOct 9, 2026·3 min03Kolmogorov-Arnold Networks Match MLP Performance With 40% Fewer SamplesOct 9, 2026·2 min04IKNO for Navier–Stokes Stays Accurate Under Rotation, With 5.6x Fewer Parameters Than TFNOct 9, 2026·2 min05LoopOPD Teaches Looped Language Models to Improve Through Their Own LoopsOct 9, 2026·2 min
MoE Router Telemetry Leaks Fine-Tuning Data: Attack Gains 2.7–9.4 pp in Accuracy
Research·arXiv cs.LG

MoE Router Telemetry Leaks Fine-Tuning Data: Attack Gains 2.7–9.4 pp in Accuracy

Oct 9, 2026·3 min
SPERA Trained on 80,000 Hours of EEG From 29,048 People and Tested on 9 Tasks
Medicine·arXiv cs.LG

SPERA Trained on 80,000 Hours of EEG From 29,048 People and Tested on 9 Tasks

Oct 9, 2026·2 min
80 Synthetic Dialogues Are Enough to Train Activation Probes for Gemma-3-27B, but Not Everywhere
Research·arXiv cs.LG

80 Synthetic Dialogues Are Enough to Train Activation Probes for Gemma-3-27B, but Not Everywhere

Oct 9, 2026·2 min
Temporal Transformer for the CAN Bus Catches Anomalies Without Sharing Raw Data
Research·arXiv cs.LG

Temporal Transformer for the CAN Bus Catches Anomalies Without Sharing Raw Data

Oct 9, 2026·2 min
Freezing the Decoder in KV-Cache Compression Saves 3x Memory but Doesn't Improve Accuracy
Research·arXiv cs.LG

Freezing the Decoder in KV-Cache Compression Saves 3x Memory but Doesn't Improve Accuracy

An arXiv paper finds that when a KV cache is compressed with SVD, fine-tuning only the encoder matches the alternatives on quality at a third of the trainable…

Oct 9, 2026·2 min
SFT-as-Context Restores General Skills in Fine-Tuned LLMs, Within 2.2 Points
Research·arXiv cs.CL

SFT-as-Context Restores General Skills in Fine-Tuned LLMs, Within 2.2 Points

A training-free method has a base model use its SFT version's answer as context. Across 19 model pairs, general-skill losses average no more than 2.2 percentage…

Oct 9, 2026·3 min
Lapras replaces CoT with latent reasoning: up to +10.79% F1 with 23.9x fewer tokens
Research·arXiv cs.CL

Lapras replaces CoT with latent reasoning: up to +10.79% F1 with 23.9x fewer tokens

Lapras, a new post-training method, teaches time series models to reason in latent space: up to +10.79% F1 over CoT and 23.9x fewer generated tokens.

Oct 9, 2026·2 min
Doctors Don't Ask LLMs About Diagnoses: 127 833 Queries vs. 58 Benchmarks
Research·arXiv cs.CL

Doctors Don't Ask LLMs About Diagnoses: 127 833 Queries vs. 58 Benchmarks

Oct 9, 2026·2 min
FedAlphaEdit Merges Knowledge Edits from Different Organizations into One LLM Without Sharing Data
Research·arXiv cs.CL

FedAlphaEdit Merges Knowledge Edits from Different Organizations into One LLM Without Sharing Data

Oct 9, 2026·2 min
Three LLMs Follow Prompts for Stutters and Interruptions Poorly; Rules Do Better
Research·arXiv cs.CL

Three LLMs Follow Prompts for Stutters and Interruptions Poorly; Rules Do Better

Oct 9, 2026·2 min
Sociolinguistics Instead of LLM Judges: Testing User-Simulator Personas on 5 Agents
Research·arXiv cs.CL

Sociolinguistics Instead of LLM Judges: Testing User-Simulator Personas on 5 Agents

Oct 9, 2026·2 min
AI4Fire: SQL Tool Lifts LLM Accuracy on Wildfire Databases From 16% to 88%
Research·arXiv cs.CL

AI4Fire: SQL Tool Lifts LLM Accuracy on Wildfire Databases From 16% to 88%

Oct 9, 2026·2 min
LLMs Judge Translation Across 70 Language Pairs: They Beat Humans Only in Places
Research

LLMs Judge Translation Across 70 Language Pairs: They Beat Humans Only in Places

Oct 9, 2026·2 min
STI-OPD: A Probabilistic Teacher Intervention for Multi-Step Agentic Tasks
LLM

STI-OPD: A Probabilistic Teacher Intervention for Multi-Step Agentic Tasks

Oct 9, 2026·2 min
MASA Reframes LLM Sparse Attention as Matrix Approximation
Research

MASA Reframes LLM Sparse Attention as Matrix Approximation

Oct 9, 2026·2 min
Routed Sparse Autoencoders Separate Words From Intonation in Speech on WavLM and SPEAR
Research·arXiv cs.CL

Routed Sparse Autoencoders Separate Words From Intonation in Speech on WavLM and SPEAR

A new TopK autoencoder method sends speech content and voice with emotion down separate routes. Tested on two encoders, two corpora and independent probes.

Oct 9, 2026·2 min
Galahad-kv stores KV cache on NVMe and loads a block 2.8–4.3x faster than recomputing it
Research·arXiv cs.CL

Galahad-kv stores KV cache on NVMe and loads a block 2.8–4.3x faster than recomputing it

The galahad-kv package stores the KV state of 50M tokens on encrypted NVMe. On an H100, loading a block is 2.8–4.3x faster than recomputing it and uses 8.8–12.3x…

Oct 9, 2026·2 min