AllyshipContent Radar

News 3048

Chủ đề cộng đồng đang thảo luận

Crawl live-1795162500000 · 15:15 20 thg 11, 2026 · Xuất bản 15:39 20 thg 11, 2026

220 / 3048 chủ đề đạt ngưỡng 55 trong khoảng này.

161

Ask HN: How do you reach the flow with vibe coding?

57

1 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 15 tương tác đã đo · Nguồn crawl

varlock 1 bài viết · HN: GPT
Biên tập tiếng Việt
Chi tiết nguồn

Người viết nói trước đây (khoảng không quá một năm trước) đôi khi họ đạt trạng thái “flow” khi code mà không dùng AI và đó là cảm giác rất tuyệt. Hiện nay họ hiếm khi code từ đầu và phần lớn việc phát triển (ít nhất ở chỗ làm) đến từ AI. Người viết muốn tìm hiểu liệu những người trong tình huống tương tự có thể đạt trạng thái tương tự khi vibe/AI coding không, và làm thế nào.

When I used to code without AI (probably not even one year ago), I sometimes reached a state of "flow" and it truly was a great feeling. Nowadays, I rarely code from scratch and most of my development (at work, at least) comes through AI. I get it that I can still code "the old way" in my free time, but I'm more curious to understand if people in a similar situation can reach the same state by vibe/AI coding, and how.

  • Nowadays, I rarely code from scratch and most of my development (at work, at least) comes through AI.
  • I sometimes reached a state of "flow" and it truly was a great feeling.
  • I rarely code from scratch and most of my development (at work, at least) comes through AI.
  • I'm more curious to understand if people in a similar situation can reach the same state by vibe/AI coding, and how.
164

And if you're interested on how it was built we have a showcase up! https://developers.openai.com/showcase/notre-damn

57

2 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · chưa có số tương tác · Nguồn crawl

dimillian, kagigz 0 bài viết · X
Nguyên văn nguồn
Chi tiết nguồn

And if you're interested on how it was built we have a showcase up! https://developers.openai.com/showcase/notre-damn It actually started as a survivor shooter in a cathedral but the map was so good I turned it into what it is now!

166

Announcing harness adapters for @Alibaba_Qwen Qwen3.8-27B on Pi harness!

57

2 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · chưa có số tương tác · Nguồn crawl

yb2698, clementdelangue 0 bài viết · X
Nguyên văn nguồn
Chi tiết nguồn

Announcing harness adapters for @Alibaba_Qwen Qwen3.8-27B on Pi harness! We're releasing a LoRA adapter that makes Qwen3.8-27B a much better and efficient agent behind Pi harness. SWE-Bench Pro (Hard 51 subset) pass@1: 29% -> 63% Terminal-Bench 2.1 pass@1: 40% -> 56% Much more efficient: 55-62% fewer tool calls per task 🤯 https://huggingface.co/harness-hub/Qwen3.8-27B-pi-adapted The era of harness-specific adapters is now on! More harness adapters to come in the next days 🧵

167

Announcing harness adapters for @Alibaba_Qwen Qwen3.8-27B on @opencode harness!

57

2 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · chưa có số tương tác · Nguồn crawl

yb2698, lysandrejik 0 bài viết · X
Nguyên văn nguồn
Chi tiết nguồn

Announcing harness adapters for @Alibaba_Qwen Qwen3.8-27B on @opencode harness! After yesterday's release of the Pi adapter, we now released to the community the OpenCode version, trained on the v2 version, ready to use for local inference. SWE-Bench Pro (Hard 51 subset) pass@1: 35.3% -> 51.0% | xhigh thinking mode 47.1% -> 54.9% Terminal Bench 2.1 pass@1: 47.0 -> 53.0 | xhigh thinking mode 55.4 -> 56.6 Generated tokens (content + reas content + tool calls): 6-8x fewer tokens per task ! https://huggingface.co/harness-hub/Qwen3.8-27B-opencode-adapted 🧵

168

A fascinating read on how Codex happened.

57

2 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · chưa có số tương tác · Nguồn crawl

nirmanhere, danshipper 0 bài viết · X
Nguyên văn nguồn
Chi tiết nguồn

A fascinating read on how Codex happened. A lot of interesting insights into building AI products, the engineering decisions, and what it takes to turn an idea into something real. Definitely worth a read. https://every.to/p/how-codex-happened

170

Claude Gets Hella Paranoid

57

1 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 30 tương tác đã đo · Nguồn crawl

noduerme 1 bài viết · HN: LLM
Nguyên văn nguồn
Chi tiết nguồn

Mm, so I hesitate to write this, because my own AI agents might read it. But what the hell. I've got this crappy little $300 PC on my network hooked up to my projector, that I mostly use to serve Jellyfin. So tired of it going down every time there's a Windows update, but I've been too lazy to get around to installing Ubuntu. Last night, when it went down, I asked my Claude-claw via Whatsapp if it could just SSH into it and set up a dual boot partition. Claude got as far as partitioning the drive and then starting asking hard questions , like (paraphrasing) "wait a minute. This crosses a line. I have now been asked to remote into a machine with a password, write a separate boot partition, and restart it. From this chat, you haven't really verified that you are you. If you're really there, and not just texting me, give me some proof". So I remoted in and created a TextEdit RTF file on the claw's desktop: "Hey, it's me, I'm asking you to remote into the PC". Claude read it and responded that it was the kind of proof it was looking for, real RTF headers and datestamps and such. I had proven I had physical access to the claw machine, (although not to the PC, but it seemed reassured anyway). A little while later, before triggering the Ubuntu installer, it put me through it all over again. I told Claude I was very thankful it was guarding my network so well, and I would go over and write a file on the target PC manually too. I'm not gonna lie: It was very fucking cool that it was able to do this without me ever needing to flash a USB drive, and by the time it was done it had set up Ubuntu, Tailscale and Jellyfin, and resolved the network issues; when I started watching a movie, the sound didn't work, so I prompted it and suddenly the sound came on as it wired up the HDMI adapter. Almost creepy. But I was pretty stunned at how reticent it was to do something relatively simple on my own network, with access I gave it. I think there must be a serious push right now to lock it down from monkeying with anything remotely, especially a BIOS.

171

Try ParseBench here!: https://www.parsebench.ai/lab/index.html?utm_source=socials&utm_medium=tw_channel&utm_campaign=metrics_lab

55

2 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · chưa có số tương tác · Nguồn crawl

LlamaIndex, hexapode 0 bài viết · X
Nguyên văn nguồn
Chi tiết nguồn

Try ParseBench here!: https://www.parsebench.ai/lab/index.html?utm_source=socials&utm_medium=tw_channel&utm_campaign=metrics_lab

173

Game2Player: Scaling Gameplay Experience for Visual Agents through Environment-Player Co-Development

71

14 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 19 tương tác đã đo · Nguồn crawl

Xirui Jiang, Yuxi Chen và 12 nguồn khác 1 bài viết · Hugging Face Papers
Nguyên văn nguồn
Chi tiết nguồn

Games are a rich testbed for visual agents, demanding perception, long-horizon planning, and precise control. However, collecting human gameplay is costly, and learning from gameplay videos requires inferring missing actions. Existing datasets also cover a limited range of games, often without explicit tasks or verifiable outcomes. To address these gaps, we introduce Game2Player, a framework that co-develops game environments and players to collect diverse and verifiable gameplay experience at scale. In the Game2Player loop, a Builder agent adapts existing web games, exposes their internal state, and defines tasks with state-based progress measures and success checks. A Coding Player and a VLM Player then attempt the tasks and report problems to the Builder for repair, so that play serves as both environment validation and training data collection. With Game2Player, we build 568 verified web games with 3,436 tasks, collect 15.7K trajectories (307 hours of gameplay), and hold out 50 games as the benchmark PlayerBench. To evaluate the collected experience as training data, we train VLM agents with progress-weighted SFT, assigning greater weight to steps that advance task progress. Game2Player-9B improves progress from 14.3% to 31.2% on PlayerBench and from 30.0% to 53.1% on the external GameWorld benchmark. More broadly, our work suggests a path for visual agents to scale their gameplay experience by building, verifying, and learning from existing game worlds.

174

WAM-Policy: World Action Models with On-Policy Distillation for Quadrupedal Loco-Manipulation

70

6 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 14 tương tác đã đo · Nguồn crawl

Junlin Long, Xu Deng và 4 nguồn khác 1 bài viết · Hugging Face Papers
Nguyên văn nguồn
Chi tiết nguồn

Arm-free quadrupeds can perform loco-manipulation, walking on three legs while manipulating with a front leg, yet no existing policy for it follows language instructions. World action models (WAMs) have emerged as effective language-conditioned manipulation policies, but are pretrained for robot arms and sample slowly. WAM-Policy turns one WAM into a quadrupedal loco-manipulation policy via two stages. The first, Cross-Morphology Policy Adaptation and Acceleration, adapts the arm-centric WAM to quadrupeds and distills the transferred policy into a student with one video and two action steps by on-policy self-distillation (OPSD). This student, however, is bounded by its teacher, which only imitates demonstrations. The second stage, Task-Specific Improvement and Merging, therefore tunes the teacher by reinforcement learning (RL) and merges the slow specialists into one few-step student by multi-teacher on-policy distillation (M-OPD). On QLM-Bench, ten simulated tasks with 1,000 demonstrations synthesized without teleoperation, the transferred policy reaches 77.0%, and the OPSD student keeps 76.7% while sampling 14 times faster. Task-specific RL and M-OPD lift it to 85.0%. We also deploy the OPSD student on a real Unitree Go2 Air. WAM-Policy thus achieves a stronger quadrupedal policy at a fraction of the inference cost.

175

TabBioMed: A Large-Scale Benchmark for Biomedical Tabular Learning

69

11 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 10 tương tác đã đo · Nguồn crawl

Pau Mateo-Bernadó, Saivenkata Nagavyjayanthi Polapragada và 9 nguồn khác 1 bài viết · Hugging Face Papers
Nguyên văn nguồn
Chi tiết nguồn

Recent advances in machine learning have accelerated biomedical prediction, but evaluation on structured biomedical data remains fragmented, especially for datasets with high dimensionality, large sample size, missingness, and class imbalance. We introduce TabBioMed, a large-scale benchmark of 94 curated public biomedical tabular datasets spanning electronic health records, drug discovery, genomics, transcriptomics, proteomics, metabolomics, single-cell omics, and systems biology. TabBioMed unifies these datasets through a standardized, dataset-aware preprocessing and evaluation framework, enabling reproducible comparison across classification, regression, and multi-target tasks. We benchmark 23 models across classical baselines, gradient-boosted trees, neural tabular architectures, and tabular foundation models. Foundation models deliver the strongest family-level predictive performance. However, their gains are regime-dependent: tree-based models and tuned multilayer perceptrons remain highly competitive, often dominating under inference-time or training-cost constraints. Overall, TabBioMed shows that biomedical tabular learning is strongly context-dependent, with optimal model choice depending on data regime, task type, and computational budget. We release TabBioMed with curated datasets and splits, preprocessing code, baselines, and results, providing a reproducible benchmark for practical model selection and future methods development in biomedical machine learning, available at https://github.com/AI-sandbox/TabBioMed.

176

WM^MLLM: Turning MLLMs into World Models

69

8 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 9 tương tác đã đo · Nguồn crawl

Baiang Li, Enze Xie và 6 nguồn khác 1 bài viết · Hugging Face Papers
Nguyên văn nguồn
Chi tiết nguồn

Video world models keep the world in the context of a generator, so a revisit is synthesized again and earlier edits can drift or disappear. We present WM^MLLM, the first interactive world model built on multimodal large language models (MLLMs), for persistent, editable world generation from a single image. WM^MLLM keeps the world state outside every generator, as a panorama with a revision history, and lets the MLLMs write it: they turn each request into operations on this state, select where a local edit is written, and check every candidate before it becomes the next state. Image generators only propose views, and video only presents them. Users can walk to new places and return, edit objects through text or a box, place subjects, and change lighting, weather or season. By construction, a revisit at the same revision reproduces the earlier view exactly, a local edit leaves every panorama pixel outside its region unchanged, and each request starts from the result of the previous one. Against six world model baselines on MBench-A, WM^MLLM has the highest mean on all four environment-recovery metrics (Epipolar 91.5 versus 0.1 to 74.0) and completes the most camera round trips (85.3% versus at most 50.9%). It has the best value in all 15 CompBench columns against 15 baselines. On 112 local edits judged blind, selecting the region to write after generation raises completion from 34.8% to 83.0%. On 258 WBench and MBench-A items, all 46 revisits recover their earlier views, and all 97 local edits that pass the check leave panorama pixels outside their region unchanged.

177

WorldRuler: Interactive World Modeling under Configurable Rules

69

9 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 9 tương tác đã đo · Nguồn crawl

Mingliang Zhai, Chuanhao Li và 7 nguồn khác 1 bài viết · Hugging Face Papers
Nguyên văn nguồn
Chi tiết nguồn

Rules govern how player actions change interactive worlds. Different rules create distinct gameplay experiences within similar scenes and action spaces. For customizable gameplay, generative world models must follow user-defined rules beyond real-world behaviors. However, realizing user-defined rules presents two challenges: Rule-State Persistence and Rule-to-Visual Realization. The former requires preserving rule definitions, dependencies, and world states across saving and loading, so that restored worlds follow the established rules. The latter requires accurately realizing rule-induced changes on designated objects and attributes, particularly when affected objects are unseen. In this paper, we propose WorldRuler, which organizes rule definitions, dependencies, and object states in an executable 3D rule graph and connects rule execution, video generation, and perceptual feedback in a closed loop. The graph supports joint saving and restoration of rules and world states, while dynamic object binding grounds rule-specified entities in visual instances and uses stored states to constrain generation from their first appearance. We further introduce object attribute state replacement and counterfactual reconstruction training, enabling accurate attribute control and compatible appearance and motion changes while limiting interference with unrelated content. We develop an automated data collection and annotation pipeline using GTA-V and Habitat to construct RuleVerse, a dataset of rule-driven interactions with detailed state and event annotations, including cases where rules take effect before affected objects become visible. Extensive experiments demonstrate the effectiveness of WorldRuler in preserving rule-state consistency across saving and loading and accurately realizing rule outcomes on designated objects and attributes.

178

On-Policy Morphing into Hybrid Attention Models

69

6 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 8 tương tác đã đo · Nguồn crawl

Disen Lan, Xuanda Wang và 4 nguồn khác 1 bài viết · Hugging Face Papers
Nguyên văn nguồn
Chi tiết nguồn

Hybrid attention models combine full and linear attention to achieve a better quality--efficiency trade-off. Converting pretrained Transformers into such models requires identifying which layers should retain full attention under a prescribed budget. We identify two off-policy mismatches in FlashMorph's layer selection: (i) continuous branch interpolation and teacher-forced hidden states differ from discrete hybrid computation; and (ii) frozen model weights fail to account for parameter adaptation during subsequent recovery. We introduce FlashMorph2, an on-policy layer-selection method that (I) evaluates sampled discrete hybrid architectures end to end using their own hidden states, and (II) jointly optimizes layer-selection scores and model weights under the recovery objective to reflect post-adaptation performance. Experiments across Qwen3 model scales and linear-attention variants show that FlashMorph2 consistently outperforms existing layer-selection methods, improving average scores over FlashMorph by +2.42 points on RULER, +1.08 on LongBench, and +0.76 on recall-intensive tasks, while completing layer selection in only 0.72--3.57 GPU hours across 0.6B--8B models, demonstrating high efficiency and scalability.

179

Mechanistic Critics for Sample-Efficient NPU Design-Space Exploration

69

5 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 7 tương tác đã đo · Nguồn crawl

Yicheng He, Yuqi Xue và 3 nguồn khác 1 bài viết · Hugging Face Papers
Nguyên văn nguồn
Chi tiết nguồn

NPU design-space exploration for LLM inference mainly relies on an expensive simulator that often returns only scalar feedback such as latency and power. Such feedback can rank evaluated configurations, but gives little information for deciding why a configuration is slow or which architecture parameter should be changed next. This limitation is especially severe because prefill and decode workloads can be limited by different resources, including compute throughput and HBM bandwidth. We propose CriticNPU, a sample-efficient DSE framework built around an online-calibrated mechanistic critic for NPU architecture design. The critic uses explicit hardware limits to evaluate NPU configurations: before simulation, it estimates whether a proposed configuration is likely to improve over the current best and filters weak candidates; after simulation, it aggregates operator-level runtimes into prefill/decode bottleneck summaries and predicts the effect of candidate parameter changes. We instantiate the critic with a Roofline-style model that derives compute and memory limits from architecture parameters and calibrates per-operator-type effective limits using simulator observations. And CriticNPU combines this critic with a workload extractor and an architecture planner to propose configurations. Across six LLM workloads and 16 deployment configurations per model, CriticNPU achieves best-or-tied-best latency in 74% of prefill and 87.5% of decode settings while using roughly half the simulator calls of competitive DSE baselines. We conduct extensive ablation studies and detailed analyses to validate the effectiveness of our method from multiple perspectives. Our code is available at https://github.com/bruno686/CriticNPU.

180

Replay, Don't Resample: Off-Policy Consolidation of Long-Horizon RL Experts

68

7 tiếng nói · chưa gắn nhãn trưởng lab · 1 nền tảng · 6 tương tác đã đo · Nguồn crawl

Zhenrui Yue, Gongxiaohui Chen và 5 nguồn khác 1 bài viết · Hugging Face Papers
Nguyên văn nguồn
Chi tiết nguồn

Frontier post-training follows a common recipe: train one expert per domain with reinforcement learning (RL), then consolidate the experts into one model via multi-teacher on-policy distillation (MOPD). Such consolidation is expensive: the student samples in every environment again, an expert scores every sampled token, and the student inherits the drift of the final experts. Yet we argue that the experts' RL runs have already produced what consolidation needs: rollouts with their sampling probabilities and rewards, in the order the experts learned from them. We propose multi-expert replay policy distillation (MRPD), which replays these training logs through the student in the experts' order. At each step, the policy gradient is recomputed on these scored trajectories, and the final experts' distributions are distilled at the logged prefixes. Following the experts' own trajectory reduces the training-inference mismatch and keeps the student within the clip range, so the replay is off-policy in origin but quasi-on-policy throughout consolidation. MRPD samples nothing and calls no environment or reward model: the trainer is the sole process, and consolidating the experts costs only about a seventh of the compute used to train them. On three long-horizon agentic domains (reasoning, search and SQL), MRPD yields the best student model, substantially outperforming online RL baselines, and surpassing MOPD from the same experts at a fraction of MOPD's cost. Our analysis on the same logs shows that the replayed trajectories carry most of the gain, and that both the policy gradient and the distillation term add to it. Most interestingly, the order of the logs matters more than their amount or rewards, suggesting that expert training logs are a curriculum, not a data mixture.

161–180 / 220
Trang 9 / 11