Archive
新模型、開放權重、多模態、語音、影像與模型架構。
代理工具
Hugging Face and Voice Arena have expanded the open speech recognition leaderboard with the Monsoon dataset. Four public and private splits cover 4,888 speakers. For Hindi, reference lattices that accept multiple spellings replace a single canonical transcription, preventing conventional WER from counting correct recognitions as errors.
最新模型
FastVideo combines DMD2 distillation with 90% sparse attention to reduce the base model’s 49 DiT calls to four, while releasing the full weights and LoRA. Its real-time generation figures depend on eight B200 GPUs, specialized VSA kernels, and fully warmed-up tests; a single GPU and mainstream consumer GPUs have yet to achieve comparable speeds.
AI 安全
Researchers used Python module shadowing to make a security decoder written by Claude Code load a malicious `struct.py`, ultimately achieving remote code execution. The case shows that per-tool-call classifiers cannot replace operating system isolation, and that a zero percent success rate in official evaluations cannot be extrapolated to unseen attack chains.
代理工具
The Model Hardware Standard packages instrument capabilities, natural-language descriptions, and safety constraints into discoverable driver interfaces, enabling agents to coordinate multiple devices through MCP, a CLI, or APIs. It remains an application-only research preview, with the specification, code, and cross-model safety evaluations not yet public.
AI 研究
CritICL first catalogs small-model errors, failure types, and critiques offline, then adds relevant examples to a large model’s prompt instead of repeatedly sampling for every question. The authors report slightly higher accuracy than self-consistency on Qwen2.5 math benchmarks, but the cost of building the library offline and the unreleased code limit reproducibility.
代理工具
Experiential unifies cloud, bring-your-own-key, and local models behind an OpenAI-compatible interface, with identity scoping, model allowlists, and hard budget controls. It can also build task routers from OpenTelemetry agent traces, although its cost-savings figures currently come primarily from the team’s own evaluations.
代理工具
The new release adds the `on_mcp_tool_result` lifecycle interface, allowing plugins to inspect or replace successful and error results from MCP. This provides a new control point for data masking and output normalization, while also elevating plugins into an agent trust boundary that requires auditing.
代理工具
PILOT connects a supervisor and a worker agent through a bidirectional channel, allowing the former to provide guidance or terminate an erroneous path before a long-running task is complete. The authors report improved accuracy and token efficiency across three agent benchmarks, but its “self-improvement” updates only the harness, skills, and memory—not the model parameters.
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.
最新模型
Puro-2B integrates a two-stage data curriculum, blockwise FP8, and MuonH optimization on a consumer-grade GPU cluster. The authors report that a shortened $4,370 training run already surpasses Qwen2-1.5B in average score, although the cost accounts only for accelerator rental-equivalent value.
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.