The Agent Architect #43: Working Memory Patterns

Pattern der Woche

Working Memory Patterns

Was:

Hält während der Problemlösung nur eine begrenzte Menge aktiver Fakten und Ziele vor und verwirft oder archiviert weniger wichtige Elemente, sobald die Kapazität ausgeschöpft ist.

Wann einsetzen:

Mehrstufige Aufgaben, bei denen der Agent Zwischenergebnisse, aktuelle Ziele und Randbedingungen verfolgen muss, ohne in jedem Schritt den gesamten Kontext erneut zu lesen.

Achtung:

Agenten behalten oft zu viel oder das Falsche, sodass spätere Schritte auf veraltete Daten zugreifen oder entscheidende Randbedingungen verlieren.

Diese Woche in agentischer KI

• FuTransformer: FP8 prefill and INT4 decode for Llama-style inference on HopperarXiv cs.AI

  Transformer prefill reuses weights across tokens and is dominated by matrix multiplication and activation data movement; single-token decode has limited reuse and is dominated by weight and KV-cache reads. We present FUTRANSFORMER, a CUDA implementation of a Llama-style Transformer block for NVIDIA Hopper GPUs. It applies CODA's GEMM-plus-epilogue approach to inference, adding direct KV-cache placement, folding the fixed RMSNorm gain into weights, writing FP8 output directly with delayed scaling, and supporting prefill without padding. Decode uses INT4 projection weights and an INT4 KV-cache c

• Beyond the Benchmark: Conditioning, Domains and Failure Modes of Qwen3-TTSarXiv cs.AI

  Qwen3-TTS reports state-of-the-art zero-shot voice cloning, but its technical report leaves open how the model should be conditioned, how it behaves outside read speech, and how it fails. We characterise its two cloning interfaces, in-context learning (ICL) and X-Vector, against Chatterbox, XTTS-v2 and F5-TTS on read, meeting, emotional and air-traffic-control (ATC) speech, using eighteen objective measures, an LLM judge, a spoofing countermeasure and a listening test. Neither interface dominates: ICL clones timbre more closely from correctly transcribed references, X-Vector clones timing, and

• Scaling Kimi K3 on AMD Instinct MI355X: Matching NVIDIA HGX B300 Throughput per GPU on AgentXvLLM

  Inside the Kimi K3 optimization effort: understanding the bottlenecks and the decisions that made inference faster

• Hyper Optimizing Kimi K3 in vLLM: Moving Fast and ResponsiblyvLLM

  How agent-assisted CuTe DSL optimizations accelerate Kimi K3 decode on NVIDIA GB300, and how scoped changes, correctness tests, and reproducible benchmarks help ship them responsibly.

• NVIDIA Nemotron 3.5 Super VL Powers Specialized Agents with Multimodal ReasoningNVIDIA Developer

  Always-on agents are taking on longer, more complex enterprise workflows. A single request can involve reading a document, checking information across sources,... Always-on agents are taking on longer, more complex enterprise workflows. A single request can involve reading a document, checking information across sources, following company policy, and calling several tools. NVIDIA Nemotron 3.5 Super VL is an open, customizable multimodal model for these specialized workflows. Its 120B-parameter mixture-of-experts (MoE) architecture activates 12B… Source