Llama 3.1 8B (AI model): news timeline & CVEs

Researchers bypass 6 AI agent frameworks with a simple ‘Repeat-After-Me’ prompt

Researchers used a Repeat-After-Me prompt to make six AI agent frameworks run attacker-written tool calls.

Researchers from Pennsylvania State University, Université Montpellier, and Dartmouth College showed a Repeat-After-Me prompt can make six agent frameworks execute attacker-written tool calls. The affected projects were LangGraph, Microsoft Agent Framework, AutoGen, Google Agent Development Kit 2.0, AWS Strands Agents, and CrewAI. In a LangGraph 1.2.0 test with Llama 3.1 8B, the orchestrator dispatched a structured call despite a refusal instruction. Their ToFu fuzzer confirmed 253 bugs in 118 of 273 MCP servers, including path traversal, SSRF, data exfiltration, and command injection; in-the-wild exploitation was not reported.

CSO OnlineOct 12, 12:07 UTC · 4d agoAI safety & security11· 4 reads

Accelerating Sharded Data Parallelism at Scale with Federated Learning

Hybrid FL+FSDP and FL+HSDP algorithms cut communication overhead in sharded data parallelism, accelerating Llama3.1 8B pre-training on 512 A100 GPUs by up to 8x.

The paper introduces FL+FSDP and FL+HSDP, hybrid algorithms that interleave sharded data parallelism with FedAvg-style federated aggregations to decouple large DP deployments into loosely-coupled federation groups. Formal communication-cost analysis and experiments on multi-tier GPU interconnects demonstrate scalability and flexibility. A Llama3.1 8B pre-training run on 512 A100 GPUs achieves up to 8.04x faster data processing and 4.48 lower evaluation perplexity than sharded DP baselines under identical hyperparameters.

arXiv cs.AI / cs.LG / cs.CLSep 17, 13:21 UTC · 29d agoAI research

Llama 3.1 8B

Timeline2 articles

Appears with

Researchers bypass 6 AI agent frameworks with a simple ‘Repeat-After-Me’ prompt

Accelerating Sharded Data Parallelism at Scale with Federated Learning