Explore tools, prompts, skills, news and resources related to TPU 第5页.
Prompt works related to TPU 第5页
Latest articles, prompts, use cases and editorial content about TPU 第5页
Simon Willison transcribed the compute section of Anthropic's IPO roadshow video, in which co-founder Tom Brown says Anthropic now trains and serves Claude across all four major chip families: AWS Trainium, Google T...
Researchers propose ReFrain (Regression-Free Abstention), which combines a linear probe on first-token logits, threshold calibration on held-out image clusters, and a deterministic veto that replaces rejected outputs wit...
Researchers propose BARS, a sequential sampling rule for majority identification under an unknown distribution with a fixed budget and target error rate. BARS certifies the top-two gap and returns either an early stop, a...
This arXiv paper studies whether key-value caches computed by one model version can be reused after weights are updated, examining four model updates with a detailed case study of DeepSeek-V4-Flash to Flash-0731. The aut...
Mercor raised a $750M Series D at a $20B valuation after reaching $3B in annualized revenue, according to the AINews roundup. The item also covers Google's Gemini 4 Carbon launch for long-horizon agentic work, Anthr...
MBZUAI’s Institute of Foundation Models released PAN 2, an open world model that takes a scene and text action and simulates the outcome as video. It uses an LLM planner with a 29B MMDiT and a frozen Qwen3.5-9B text enco...
This arXiv paper studies fingerprinting agentic cyberattacks by inferring the attacker's LLM version and harness from telemetry. The authors introduce "shibboleth traps" — modular honeypot services that co...
DForkTree is a training-free speculative decoding method that re-anchors block-diffusion draft-tree branches on their own tokens. It adds n-gram pointwise mutual information from the target model's outputs to drafte...
Moonshot AI's Kimi K3.1 Pro Highspeed is now available on Amazon Bedrock. It is a post-trained version of Kimi K3 with the same 3T-class open-weight architecture and 1-million-token context window, trained with more...
PARIS is an open-source framework in which a conductor agent running in Claude Code orchestrates 15 specialized LLM agents across a systematic review workflow, from research scoping to manuscript draft export. Agents coo...
AWS published a playbook for evaluating production AI agents on Amazon Bedrock AgentCore Evaluations. It describes built-in AWS evaluators, managed DeepEval and AutoEval metrics, and custom LLM-as-a-judge evaluators, inc...
AWS describes the now-available Web Search API in Amazon Bedrock AgentCore and its LangChain integration, which returns current, cited web results using existing AWS credentials with configurable domain and date filters....
A new arXiv paper presents two empirical studies estimating the energy and carbon footprint of agentic language-model systems. In iterative self-refinement on MATH, three refinement steps closed a 16-percentage-point acc...
Latent reviewed NaiveAI's Naive-N0.5-Flash, an open-weights coding model derived from Xiaomi's MiMo-V2.5 and paired with the company's NaiveRT inference engine. NaiveAI claims peak throughput around 2,100...
Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2, describing them as the same underlying frontier model offered at two safeguard levels. Fable 5.2 keeps classifiers that block high-risk cybersecurity and most re...
A new arXiv study finds that storing cached recurrent state in bf16 with round-to-nearest makes hybrid reasoning models more verbose and increases 32K-token cap hits. Across six open hybrid models and 590 competition-mat...
The paper presents an evaluation framework for AI agents doing chip floorplanning that connects submitted layouts to a prescribed downstream implementation flow, checks geometric validity, measures physical quality and r...
A controlled study with 20 novice programmers found that using generative AI while reading code increased objective cognitive workload by 11.2% in reaction time, even though participants perceived no extra effort. AI ass...
Meta AI launched Muse Ember, described as its smallest and most efficient model, targeting complex reasoning execution and repetitive agent work such as tool calls, coding subtasks, routing, extraction, and long document...