AI Coding Live News

On 2026-11-06 Reflection's official account released the Beam model technical report (a 156-page PDF), detailing how its 501.3B-total / 23.4B-active open-weight MoE was trained: 23.8T pretraining tokens on 6,144 GB300 GPUs in under four weeks, then a fully asynchronous RL run on about 10.5K GB300s for four weeks that produced over 100M rollouts. The report says FP8 and NVFP4 checkpoints keep 99.93% and 98.98% of median benchmark scores, and the 1.93B DFlash 2 drafter speeds single-request decoding 1.96–3.03x (vendor figures).

Zoom

Researchers from the University of Maryland, Google, and MBZUAI introduced BeliefWeaver (arXiv:2610.18082), an autonomous failure-recovery framework for long-horizon agents. Addressing cases where failures stem from root errors multiple steps earlier that local reflection cannot fix, BeliefWeaver maintains a persistent relational belief state across checkpoints and attempts, using an Implicit World Model (IWM) to evaluate hypotheses and gate repair execution. When distilled into a 9B model via on-policy distillation, it improves final benchmark scores by 9.8 percentage points while slashing teacher API costs by 54.8%.

Zoom

Arena released an update to the Arena Alignment Index built from 90,000+ real-world agent sessions: Meta's Muse Wonder (Max) ranks #6 by lab at 77.5% (2.09% unauthorized action rate, 14.41% deceptive completion rate), Reflection AI's open-weight Beam-1.0-501B-A23B debuts at #7 with 75.3% (4.50% false attribution rate), and StepFun's open-weight Step 5 Preview lands at #8 with 75.1% (2.11% unauthorized action rate).

Zoom

Google AI Studio launched Gemini 4 Argon access in AI Studio Build and Playground for Google AI Pro subscribers, while Hugging Face Transformers v5.20.0 and v5.20.1 added native support for Reflection AI's frontier coding model Beam, Meta's 816B-parameter (42B active, 1M context) multimodal agentic coding model Muse Spark Open, InclusionAI's Ling 3.0 Flash, and OpenBMB's MiniCPM-V 4.7.

Zoom

OpenCode (@opencode), the wildly popular open-source terminal coding agent, announced free access to both Zhipu AI's high-speed GLM 5.3 Flash and Reflection AI's newly released 501B/23B MoE model Beam with a full 1M-token context window. The GLM 5.3 Flash announcement alone surged past 1.92 million views and 17,500 likes on X within hours.

Zoom

Whether LLM-based autonomous agents truly learn from interactive experience or merely regurgitate memorized knowledge remains a fundamental question in AI research. Most conventional agent benchmarks evaluate tasks (such as writing standard Python or navigating operating systems) heavily represented in pretraining corpora, creating severe data contamination that obscures whether agents are synthesizing new concepts or retrieving memorized patterns. Researchers introduce Learn2Play-Bench (arXiv:2610.08215, code: github.com/liushiliushi/Learn2Play-Bench, site: liushiliushi.github.io/learn2play-bench-website), a novel open-source benchmark constructed around text-based environments with entirely unfamiliar and counterintuitive mechanics. Agents must discover game physics and causal dependencies purely through trial-and-error interactions. The benchmark reveals two critical insights: retaining uncompressed, verbatim interaction logs decisively outperforms LLM-generated rule summaries, and the agent harness scaffolding (managing tool loops, memory, and reflection) exerts a larger impact on final performance than the choice of underlying foundation model.

Zoom

As autonomous LLM agents navigate complex long-horizon operational streams in production, enabling continuous test-time weight adaptation without catastrophic policy drift remains a foundational challenge. Conventional in-context adaptation stores reflections or skills as text, making reuse bottlenecked by fragile retrieval heuristics over a static, frozen policy; conversely, directly imitating or reinforcing tokens from single deployment rollouts destabilizes policy weights. Researchers formalize Online Agentic Test-Time Training (OaTTT) and introduce ASCENT (arXiv:2610.05303). The agent executes each task in a single streaming pass. A frozen copy of the initial model acts as a hindsight teacher, receiving verified execution trajectories as privileged context to compute calibrated next-token predictive distributions. Distilling these distributions into persistent LoRA fast weights continuously updates policy parameters across subsequent tasks without external teacher models or gold references. By pruning invalid-action turns to distill condensed execution trajectories, ASCENT systematically improves task success rates and interaction efficiency across ALFWorld, WebShop, and AppWorld, successfully transferring learned behaviors to held-out environments.

Zoom

On 2026-10-16 Hugging Face and Reflection confirmed Beam weights are downloadable: ~501B-total / ~23B-active sparse MoE, 1M context, Apache 2.0. BF16/FP8/NVFP4 checkpoints ship with day-0 transformers≥5.20.1, vLLM and SGLang. Model-card figures (vendor): SWE-bench Verified 81.6% (BF16), Terminal-Bench 2.1 80.1%, MCP Atlas 80.8%. This closes the Oct 5 promise to release weights later that month.

Zoom

Test-time self-improvement—the ability of an agent to iteratively refine solutions during inference—relies on high-fidelity reflection to surpass current proposals and long-horizon execution to maintain coherent multi-round exploration. Researchers from Renmin University of China, Microsoft Research, and USTC introduce AREX-2, establishing that these capabilities are domain-agnostic and learnable from verifiable environments. Synthesizing long-horizon reflective trajectories across machine learning and algorithmic programming, an AREX-2 agent built on Qwen3.8-27B achieves 81.8 on MLE-bench Lite and 70.7 on Frontier-CS, transferring seamlessly to deep research with 84.0 on BrowseComp, 92.2 on GAIA, and 93.8 on DeepSearchQA while monotonically scaling with compute budget.

Zoom

Endowing LLM agents with test-time self-improvement—the ability to iteratively refine solutions—is a vital milestone for autonomous engineering. Researchers from Renmin University and collaborators present AREX-2, an open-source framework leveraging long-horizon reflective trajectories. Decoupling self-improvement into reflection (generating superior alternatives) and long-horizon execution (sustaining iteration over multiple turns), AREX-2 trains on verified machine learning and algorithmic programming trajectories. Built on Qwen3.8-27B, AREX-2 achieves 81.8 on MLE-bench Lite and 70.7 on Frontier-CS, transferring effectively to deep research with 92.2 on GAIA and 93.8 on DeepSearchQA.

Zoom

While test-time compute scaling has propelled reasoning capabilities in large models, small reasoning models (sRMs, 4B-8B) frequently fail when scaling thinking tokens: self-refinement merely consolidates probability onto already reachable hypotheses, failing completely when missing parametric knowledge. KAIST researchers introduce FlyBy, a selective querying framework that formally decouples execution bottlenecks (recoverable through internal reflection) from knowledge bottlenecks (requiring external teacher queries). Optimized via cost-aware reinforcement learning, FlyBy teaches small models when to think, when to ask, and how much compute budget to expend. Evaluated across 1,158 hard problems on six benchmarks, FlyBy-4B reaches 45.96% pass@8, outperforming the much larger Qwen3-14B (41.64%) while slashing serving costs by 2.7x (a 63% reduction).

Zoom