Llama-3.1-8B

Named by 3 kept items in this workspace.

2 items name it under affected products and 1 under tools.

The items that name it 3

Anthropic tests AI autonomous replication across eight nodes in four of five trials
articleas tools2026-10-12It is written here as Llama3.1-8B (inference established on target nodes).
inference established on target nodes
Understanding the Impact of LLM Watermarking on AI Agent Behavior
writeupas affected products2026-09-26
Model tested for watermarking effects on tool calling and refusal
Sockpuppeting: How a Single Line Can Bypass LLM Safety Guardrails | TrendAI (US)
writeupas affected products
Open-weight model; ASR as high as 77% per Dotsinski & Eustratiadis paper

In every workspace you can read 16

AI

Folding Attention Biases into Keys Makes Compacted KV Caches Run on Standard Kernels
paper2026-10-27
FuTransformer: FP8 prefill and INT4 decode for Llama-style inference on Hopper
paper2026-10-16
Anthropic tests AI autonomous replication across eight nodes in four of five trials
article2026-10-12It is written here as Llama3.1-8B.
AI models may be capable of self-replicating without access to a datacenter
writeup2026-10-12It is written here as Llama3.1-8B.
DualQuant: Dual-scale Calibration-free Weight-and-Activation Quantization for LLMs
paper2026-10-09
Defense-in-Depth for LLMs: Evaluating Memory Gates Against Activation-Induced and Memory-Induced Sycophancy
paper2026-10-05It is written here as Llama 3.1 8B.
Understanding the Impact of LLM Watermarking on AI Agent Behavior
writeup2026-09-26
ClusterMAX 3.0: The Industry Standard GPU Cloud Rating System Returns
article2026-09-23It is written here as Llama 3.1 8B.
Accelerating Sharded Data Parallelism at Scale with Federated Learning
paper2026-09-17It is written here as Llama3.1 8B.
Retrochainer Post
thread2026-08-01It is written here as Llama 3.1 8B.
vLLM v0.30.0
repo
LLM Inference in a Flash!
paper
LLM Inference in a Flash! (UC Berkeley)
paper

Security

Anthropic tests AI autonomous replication across eight nodes in four of five trials
article2026-10-12It is written here as Llama3.1-8B (inference established on target nodes).
Understanding the Impact of LLM Watermarking on AI Agent Behavior
writeup2026-09-26
Sockpuppeting: How a Single Line Can Bypass LLM Safety Guardrails | TrendAI (US)
writeup

All technologies in this workspace.