AI Coding Live News

Researchers from Alibaba and collaborating academic institutions introduced FocusGUI in arXiv:2610.26203, establishing a focused attention architecture for streaming omni-modal GUI agents. Across streaming observations of continuous UI screenshots, synchronous audio, and device state histories, agents routinely suffer from 'Sequential Distraction'—where transient visual clutter from previous steps compounds into catastrophic action hallucination. FocusGUI preserves the base model while introducing action-sufficient projection and channel decomposition. Tested across 29 smartphone apps on OmniGUI, FocusGUI elevates Type Match from 31.0% to 64.2% and Exact Match from 11.2% to 33.5% across four omni-modal foundation backbones.

Zoom

Addressing KV cache bottlenecks in long-context LLM serving, a new production study on 8xB300 nodes serving GLM-5.3 (arXiv:2610.21500) reveals that unsharded remote KV caches inflate hit-heavy p99 TTFT to 83.32s, and resolves it via rank-sharded asynchronous LMCache loading—boosting GET bandwidth 6.3x to 10.76 GB/s and slashing p99 to 0.67s (+34% throughput). Meanwhile, Baidu PaddlePaddle open-sources Virtual Head Attention (VHA, arXiv:2610.17959), folding linear Q-Premix and Postmix into standard GQA-2 weights at inference to halve KV cache while beating MLA (+0.70%) and GQA-4 (+2.60%) on a 30B-A3B MoE and lifting 8K-128K RULER from 73.0 to 84.4.

Zoom

Off-policy data in reinforcement learning with verifiable rewards (RLVR) causes severe policy gradient bias during asynchronous generation and response reuse. Researchers from UIUC and HKUST introduce Proximal KL, theoretically proving that full future KL credit renders the exact proximal optimizer the unique stationary policy without requiring full-response weights or fitted critics. Evaluated across Qwen3-1.7B and 4B models, Proximal KL outperforms GRPO on 8 math benchmarks by up to 2.58% while completely averting training collapse.

Zoom

Nous Research shipped Hermes Agent v0.22.0, "The Threshold Release", on Oct 14 (UTC). Fresh desktop installs open a guided setup chat on a free trial model (nous/welcome) with no API key, connectors reach Gmail, Linear and Notion through the Nous tool gateway, a one-click llama.cpp runtime runs local models, installs now follow tested SemVer stable releases, and desktop builds are signed (including the Microsoft Store). The source tree shrank from about 1.06M to 698K lines (−34%).

Zoom

In software engineering and knowledge-work tasks, autonomous coding agents operate via Direct Workspace Interaction (DWI)—issuing shell commands to navigate, inspect, and modify files within local repositories without external vector indexing. However, during extended interaction horizons, agents suffer severe context degradation: initial task requirements are purged from context windows by voluminous file dumps, read file provenances vanish, and listed but unopened candidate files are permanently forgotten, leading to premature task terminations and unfulfilled requirements. Researchers from KAIST introduce RunningTab (arXiv:2610.10444), an innovative agent interaction framework that replaces unreliable internal agent scratchpads with an environment-maintained 'running tab'. The tab persists three critical streams: dynamic requirement trackers, verified file excerpts with provenance links, and unopened candidate file stacks. When an agent attempts to finalize a deliverable, the environment triggers a mandatory 'Finish Check' barrier that blocks completion if unsatisfied criteria persist. Evaluated across Workspace-Bench, TheAgentCompany, and OfficeQA Pro benchmarks using frontier LLMs, RunningTab decisively outperforms vanilla DWI and self-refine baselines, establishing substantial gains in rubric satisfaction (TCR@70).

Zoom

Token-level LLM routing—routing easy tokens to small models and switching to frontier models for critical reasoning tokens—presents an optimal trade-off between output quality and inference cost. However, current LLM serving engines (such as vLLM and TGI) are architected for single-model execution, suffering severe step desynchronization, batch admission latency, and extreme orchestration complexity when handling high-frequency token transfers across heterogeneous models. Researchers from Tsinghua University's NICS Lab present TokenRouter (arXiv:2610.12242, accepted at NeurIPS 2026, code: github.com/thu-nics/TokenRouter), an efficient and developer-friendly serving system engineered for token-level routed LLM inference. TokenRouter introduces a 'request-centric programming and model-centric execution' paradigm, enabling developers to write routing logic intuitively per request while the runtime handles asynchronous execution across subservers. Incorporating a Delayed-Batching Scheduler parameterized via discrete-time Markov chain modeling, a decoupled tri-loop architecture, and an instant handoff-resume mechanism, TokenRouter achieves 2.01x to 64.15x higher decoding throughput compared to existing serving platforms across diverse routing workloads, with complete code open-sourced.

Zoom

On Oct 7 Nous Research, maker of the open-source Hermes Agent, confirmed a $90M Series B at a $1.5B valuation led by Robot Ventures, with NVIDIA, Microsoft's M12, Samsung, USV, Y Combinator and Menlo among backers, bringing total funding to $158M. Nous says Hermes Agent has been cloned over 24M times and drives about 2.5% of global token usage (internal estimate). The money funds Hermes for Businesses: a team tier with shared balance and skill library, and an Enterprise edition on customer-controlled infrastructure with SSO and SLAs, plus a mobile app.

Zoom

Autonomous robot manipulation has long relied on specialized vision-language-action (VLA) models whose rigid training precludes zero-shot adaptation to novel environments, cutting them off from rapid foundation VLM advancements. Conversely, recent robotic agents rely heavily on fragile coding agents and auxiliary grounding models like SAM3, introducing severe latency and operational brittleness. Researchers from UIUC and CMU present MotorMind (arXiv:2609.38078), a direct robotic harness connecting general-purpose VLMs to deterministic robot control. By translating VLM mid-level actions into deterministic Cartesian primitives paired with asynchronous execution monitoring and background memory updates, MotorMind achieves 66.7% zero-shot success on LIBERO-PRO suites (versus 13.3% for prior baselines) and achieves 95% average success on a physical xArm6 robot under live human disturbances.

Zoom

Large language model serving exhibits bifurcated hardware constraints: prefill is compute-bound, benefiting from high-throughput low-precision tensor operations (e.g., NVFP4), while decode is memory-bandwidth-bound, requiring ultra-compact weight storage (e.g., 2-3 bit weight-only) to minimize token-generation memory traffic. Uniform quantization forces suboptimal compromises across both phases. Researchers from IST Austria introduce Disaggregated Quantization (DQ, arXiv:2609.26333), decoupling computation formats, quantization precision, and physical memory placement across prefill and decode. Paired with Offloaded Disaggregated Prefill (ODP) to stream NVFP4 prefill weights asynchronously from NVMe SSDs, DQ delivers a 1.78x time-to-first-token (TTFT) speedup on Qwen3.8-27B at 8K prompt length within llama.cpp, while supporting disaggregated serving in vLLM on architectures up to 2.8T parameters.

Zoom

Mainstream LLM agents strictly follow sequential interaction loops: read, think, reply, call tools, and block until completion. Real-world applications—including full-duplex voice assistants, robotic manipulation, and continuous infrastructure monitoring—are inherently asynchronous, receiving incoming observations while the agent is actively deliberating or waiting on tool I/O. Researchers introduce a general asynchronous LLM framework that abstracts model inference into non-blocking coroutines with overlapping memory states. Without requiring task-specific fine-tuning or architectural alterations, foundation models like Qwen 3.x exhibit native asynchronous competency, effortlessly handling live video feeds, real-time gaming, and continuous systems monitoring.

Zoom

While looped language models enable depth-adaptive inference by executing fewer layer repetitions on easy tokens and more on complex tokens, variable loop counts break conventional batching engines like vLLM. Researchers from TUM and Imperial College introduced Continuous Depth Batching (CDB). By dynamically reorganizing batches between loop iterations, orchestrating looped KV-caches, and asynchronously predicting token exits, CDB achieves up to 99% of the theoretical maximum inference speedup across Ouro 1.4B and Huginn 3.5B architectures.

Zoom