Archive
GitHub 熱門專案、開源工具、框架、runtime 與重要 release。
最新模型
CLAP first learns physical dynamics from unlabeled videos through latent actions, then maps the representation back to robot end-effector coordinates. The research team has released the code and multiple checkpoints, but physical hallucinations in generated videos still limit its direct use in safety-critical control.
代理工具
PRAXIST has multiple research agents propose changes over successive generations, then records evaluation results, mechanisms, and lineage in an auditable evidence graph to avoid restarting exploration in every round. Its authors report 60 medals across 75 MLE-bench tasks at a model cost of $3,054, but the newly released system’s performance and costs have so far been measured primarily by its development team.
最新模型
Rather than treating the majority-vote answer as ground truth, TTPO distills trajectories that agree with the consensus and uses grouped reinforcement learning to penalize high-confidence disagreements. The authors report that Qwen3-1.7B’s Avg@12 across three mathematics competitions rises from 38.0 to 45.2, though this remains a transductive setting in which parameters are updated directly on the test problems.
推論系統
During generation, Prefix Sliding evicts intermediate reasoning tokens while retaining the system prompt, task prefix, and the most recent several thousand tokens. It can be applied without retraining. The research reports that overall thinking time for long reasoning can be reduced by roughly threefold, but the current implementation depends on forked versions of vLLM and FlashAttention, with limited benefits for short outputs and tasks that require revisiting early details.
最新模型
LeVJEPA uses a single encoder, a cross-view invariance loss, and SIGReg to prevent representation collapse, eliminating the need for an EMA target network, stop-gradient, or a pixel decoder. Under the same data and training schedule, the authors report a 5.6× to 20.8× compute difference, though the results remain limited to frozen-encoder benchmarks based on models retrained by the team.
推論系統
Instead of offloading MoE experts according to fixed rules, FreeToken dynamically divides work between CPU execution and GPU caching based on measured memory and PCIe bandwidth. The authors report that an RTX 5090 can run a 284B model at 22–25 tokens per second, though testing was conducted by the development team and the full weights still require substantial system memory.
硬體
Meta’s MTIA 300 integrates 1.2 TB/s of network I/O and sixteen messaging engines directly into the accelerator, preventing collective operations from consuming the main compute array. For a 150B-parameter recommendation model, Meta reports a 3.9× reduction in communication time versus a reference GPU cluster, but has yet to provide complete system data sufficient to compare total training costs.
AI 安全
Anthropic’s automated alignment research agent can independently review the literature, design data, and post-train models, closing 26% to 96% of the evaluation gap across ten safety problems. It also outperformed a constrained baseline of human proposals, though the agent received more time to iterate, and the results demonstrate only its ability to remediate measurable failures.
代理工具
The Stanford-led open benchmark packages scientific work—including data analysis, simulation, theorem proving, and instrument calibration—as reproducible terminal tasks. Claude Opus 5 paired with Claude Code ranks first, but achieves a resolution rate of only 30% across three trials, showing that high scores on general coding benchmarks cannot be directly extrapolated to scientific work.
代理工具
DeepChart evaluates data extraction, numerical derivation, and chart rendering separately across 1,482 instances, exposing data hallucinations concealed by appearance-based scores. With ultra-long 10-K inputs, the average source-data F1 score falls from 0.385 to 0.167, showing that larger context windows cannot replace evidence retrieval and localization.
Developer platforms / enterprise AI
Starting as early as September 28, GitHub plans to unify the experiences and policies for the web, mobile, and cloud agent versions, with the unified experience enabled by default. At the same time, Copilot code review’s default analysis effort will move from Lite to Balanced, significantly increasing the estimated AI credits cost per review.
AI 安全
Following the restricted mode introduced in 2.1.248, Anthropic has patched five more categories of trust-boundary issues involving files, plugins, and settings. The new release also exposes model-switch hooks and prompt cache metrics, allowing enterprise gateways to intercept switches and track recaching costs.