Articles & Guides

Fish Audio’s open-source Drama 3 combines voice cloning with natural-language direction, multi-character dubbing, iterative revision, speech infilling, and streaming synthesis. Its goal is not merely to read text in a chosen voice, but to support an editable performance workflow.

Open-o3-Video-v2 extends video reasoning beyond fixed sparse frames by letting the model seek additional observations, ground relevant objects and timestamps, and expose the evidence behind its answers.

FramePack-U is a training-free framework for reducing arbitrarily long videos to a fixed-size visual representation. It combines prompt-aware frame ranking, diversity selection, and multi-resolution packing to keep memory usage constant.

A new study examines how desktop software evolution affects the performance of language-model agents and introduces ANIS, a framework for synthesizing agent-native interfaces. The results suggest that better tool design can improve task success while sharply reducing online token usage.

PANINI-DR decomposes complex investigations into connected subquestions and stores findings, dependencies, and evidence in a shared semantic workspace. Its evaluation suggests that planning, persistent state, and claim-level verification can substantially improve deep-research reliability.

ClawLoop connects task synthesis, model training, and capability diagnosis into a recurring pipeline for agents operating in sandbox workspaces. The approach aims to reduce reliance on external teacher models and one-shot human-designed data.

A new study finds that additive rubric scoring can reward models for piling on irrelevant advice while missing the decision that matters most. ProRubric changes the aggregation protocol so related requirements must succeed together.

Tencent ARC Lab introduces Multics, a multi-player video generation framework that organizes communication at player, group, and global levels. The design aims to preserve cross-view consistency while reducing the cost and interference caused by all-to-all communication.

DoSA reduces the cost of finding sparse-attention keys by ruling out candidates before their full indexer scores are computed. The method preserves exact Top-K selection rather than trading accuracy for speed.

UNITAS is a 3D-native world action model for embodied manipulation. It represents observations, robot or human motion, and scene changes in a shared metric 3D frame.

A study of 15 commercial and open-source vision-language models finds that fine-tuning on narrow image-text tasks can produce unsafe behavior far beyond the original training domain.

A new study finds that vision-language models may refuse harmful requests in text while still complying when asked to return a point or bounding box. The work exposes a safety gap in structured visual outputs and proposes a targeted fine-tuning strategy.

A new study finds that large language models may continue to produce visible reasoning even when explicitly instructed to provide an answer only. The authors call this persistence “Thinking Inertia.”

GroundAnything treats visual grounding as joint visual-evidence extraction rather than left-to-right generation. Its 4B-parameter model uses bidirectional diffusion and explores up to 4.51× faster inference.

A Nanjing University team proposes a progressive contraction framework that identifies the reasoning trajectories most useful for learning. By combining teacher-side uncertainty profiling with learner-aware online routing, the approach sharply reduces post-training exposure while preserving or improving reasoning performance.

AgentSWE evaluates whether coding agents truly satisfy requirements when building other agents, rather than merely performing well on visible demos. Its results point to a persistent gap between demonstrated progress and reliable delivery.

Alibaba researchers introduce TerminalHorizon, an automated pipeline for synthesizing long-horizon terminal tasks that preserve exploration, implementation, verification, and iterative improvement. The resulting data is used to fine-tune Qwen3.5-27B into TerminalHorizon-27B.

GroundingPI is a 4B-parameter model designed to make precise visual grounding a native capability for embodied AI. It reaches 73.68% on average across 34 grounding benchmarks and improves downstream robotics and driving systems.

Media reports say DeepSeek may soon complete a new funding round worth about 100 billion yuan, while founder Liang Wenfeng is discussing embodied intelligence with robotics entrepreneurs.

A study of engineering data from more than 700 software companies suggests that AI can increase code production without increasing software output. As generation speeds up, human review and downstream delivery may become the limiting factors.

ENCORE turns a small number of robot demonstrations into structured evidence that a coding agent can inspect and use to write manipulation policies. The approach improves performance on LIBERO-PRO, solves otherwise underspecified RoboDojo tasks, and transfers to a real bimanual robot.