CallCleaner distills SAM-Audio into a 19M bioacoustic denoiser
Authors say the CPU model runs about 90 times faster than real time and leaves 11 dB less noise than Biodenoising.
Stories per day · 30 days
4 RSS
Relatedopenai4deepseek3api2api-keys2compaction2opencode2open-weights2reasoning-traces2ai-coding1anthropic1api-security1bioacoustics1black-box-detection1chain-of-thought1china1cifar-101credential-theft1data-breach1
24 stories, newest first · one card per story
Stories tagged #model-distillation
StepX-OPD merges distillation stages for faster diffusion models
StepX-OPD lets one student learn all capabilities and few-step sampling using its own rollout states, scoring highest among four-step methods on SD3.5-Medium and FLUX.1-dev while using 37% fewer GPU-hours.
Uncensored Heretic variant of DeepSeek-V4-Pro Qwen 9B trends on Hugging Face
The uploader says Heretic cut refusals from 100/100 to 3/100, with KL divergence of 0.0422 versus the Jackrong distill.
Chinese labs reportedly distill OpenAI GPT-6 Astra outputs
The Information says DeepSeek and Moonshot used Astra, including hidden reasoning, to train Kimi K3.1 Pro and V4.1-Flash 1017.
Researchers extract hidden GPT-5.6 reasoning through an OpenAI API flaw
OpenAI confirmed the bug, paid Xint a $6,000 bounty, and deployed an undisclosed fix on October 23.
OpenAI compaction flaw let researchers extract hidden model reasoning
Xint hid GPT-5.6 Sol reasoning in a compaction and recovered it through GPT-4.1 and GPT-4o; OpenAI paid $6000.
1
Watermark Basin Alignment detects distilled LLMs without model weights
At a 0.1% false-positive rate it detected 79% of distilled models at n-gram=1, versus 29% for a token baseline.
OpenAI API compaction flaw leaked hidden model reasoning traces
Xint says OpenAI confirmed the bug, paid $6,000, and shipped an undisclosed fix on October 23.
1
Paper recovers neural-network parameters by extraction and distillation
On the largest synthetic net, Leaky ReLU recovered 51-57 deepest-layer neurons while ReLU recovered none.
US orders civilian agencies to remove DeepSeek from systems
The DHS order also covers third-party products built with DeepSeek technology, with exceptions available on request.
OpenCode says breach hit 1.4% of Console users
The company ties the October 17 incident to a distillation attack on users with external providers, but has not confirmed stolen API keys.
2
OpenCode breach exposes API keys for 1.4% of Console users
Some linked-provider users report overnight charges up to $1,700 and Google Cloud project suspensions.
Poster: A Preliminary Study of LLM Distillation Inference
Distillation-inference test flags Llama-3.2-3B models distilled from Qwen2.5-7B at TPR 1.0 and p=0.02.
OpenAI Blocks 15,000 Requests Trying to Extract Protected Model Reasoning
OpenAI blocked over 15,000 accounts trying to extract protected model reasoning, linking much of the activity to Moonshot AI.
3
National cyber director: Government-industry collaboration vital to managing AI risks, competition with nations
National Cyber Director Sean Cairncross said industry collaboration, not heavy regulation, is key to AI security against China.
2
OpenAI says it stopped a campaign to steal its models' reasoning, but the trick still worked on Azure
OpenAI blocked a large campaign to steal model reasoning, but researchers say Azure still leaked it.
OpenAI Disrupts Reasoning Extraction Campaign Linked to Moonshot AI Associates
OpenAI disrupted a scaled reasoning-extraction campaign it links to people associated with China's Moonshot AI.
3
Irony alert: OpenAI whines that Chinese model stole its special IP that it stole from everybody else
OpenAI says it disrupted a July distillation campaign tied to Moonshot AI that copied protected model reasoning.
4
Stolen AI credentials feed growing LLM proxy economy
Team Cymru estimates more than 80,000 proxy servers hide stolen AI credentials and enable model distillation.
7
80,000 relay servers help users in China slip past U.S. AI region bans
Over 80,000 relay servers enable users in China to bypass region bans on U.S. AI models, facilitating potential model distillation and abuse.
I had Gemini train its own replacement for $9
A developer used Gemini 3.1 Pro to label 4,290 Reddit comments for $9, then fine-tuned GLiNER 459M to 0.83 F1 for product NER.
3
RAPID: A Real-Time Defense Against Unauthorized Model Distillation for Text-to-Image Services
RAPID embeds defensive perturbations in a T2I model's shared VAE decoder to block unauthorized black-box distillation in real time.
How hackers used Claude for missiles, drone swarms, and surveillance, while Chinese labs mined it for training data
Anthropic's threat report details eight months of Claude misuse: AI-assisted espionage against 20+ organizations, self-rewriting malware, and Chinese labs distilling Claude via fraudulent accounts.
US Agencies Warn China Is Systematically Extracting Frontier AI Capabilities
NSA, CISA and FBI warn Chinese AI firms including DeepSeek and Moonshot systematically extracted billions of tokens from US frontier models since late 2024.