از دل پژوهش

رصد هوش مصنوعی

هر روز، پژوهش‌های تازه و ایده‌های جذاب هوش مصنوعی را به زبان ساده مرور کنیم.

این بخش چطور آماده می‌شود؟ ←

arXiv

SIEVE: Selective Unlearning in Vision-Language Models via Attention-Value Suppression

SIEVE suppresses attention values for forget examples and matches them to a reference model for retain…

Pouya RostegarOctober 4, 2026

arXiv

HydroJEV: Fast Screening for Attribution of Water-Network Anomalies

A preprint tests the training-free Jev screen for attributing SCADA alarms in water networks; its independent…

Nazanin RadOctober 4, 2026

arXiv

A Four-Axis Taxonomy for RAG Systems

This survey organizes RAG research around efficiency, defense, interactivity, and complex reasoning, while highlighting their shared…

Arash NikrooOctober 4, 2026

arXiv

Mingbird: A Local-First Harness for Small-Model Agents

This preprint examines whether improving an agent harness, without changing the model, can make small open…

Nazanin RadOctober 4, 2026

arXiv

VISTA: Lossless Visual Memory for Multimodal Agents

VISTA is a visual harness that lets multimodal models retrieve and re-examine past frames at their…

Hossein AiOctober 4, 2026

arXiv

When the Same Reward Teaches Different Visual Skills

This preprint argues that improved visual-benchmark scores need not come from learning from images; task design…

Amir SharifiOctober 4, 2026

arXiv

Mem++: Non-Destructive Long-Term Memory for Organizational Agents

Mem++ preserves complete organizational documents with dates and authors, deferring the choice among conflicting versions until…

Leila MovahedOctober 4, 2026

arXiv

Global Coherence in Multi-Agent Collaboration

This paper studies cases where each agent acts correctly from its local view, yet the team’s…

Hossein AiOctober 4, 2026

arXiv

Sphere Encoder 2: Image Generation with Rotation in a Spherical Latent Space

This paper introduces Sphere Encoder 2, a single-stage autoencoder for image generation that improves ImageNet samples…

Mehran DadbehOctober 4, 2026

arXiv

UTF-8/UTF-16 Routing to Reduce Cross-Script Token Gaps

UBE preserves the BPE core while routing 3-byte UTF-8 characters through UTF-16 to reduce cross-script token-budget…

Nima FarahmandOctober 4, 2026

arXiv

Interactive Code Execution for Reducing Model Calls in Robot Agents

PyRUA-Lean combines Python programming, selective feedback, and persistent within-episode state, reporting higher success than a tool-calling…

Farzaneh ParsaOctober 4, 2026

arXiv

Turning 2D Vision-Language Models into 3D Programmers with Iterative Feedback

3D-Prog proposes using 2D vision-language models for 3D tasks without retraining through coordinate framing and execution…

Nima FarahmandOctober 4, 2026

رصد هوش مصنوعی

SIEVE: Selective Unlearning in Vision-Language Models via Attention-Value Suppression

SIEVE: Selective Unlearning in Vision-Language Models via Attention-Value Suppression

HydroJEV: Fast Screening for Attribution of Water-Network Anomalies

HydroJEV: Fast Screening for Attribution of Water-Network Anomalies

A Four-Axis Taxonomy for RAG Systems

A Four-Axis Taxonomy for RAG Systems

Mingbird: A Local-First Harness for Small-Model Agents

Mingbird: A Local-First Harness for Small-Model Agents

VISTA: Lossless Visual Memory for Multimodal Agents

VISTA: Lossless Visual Memory for Multimodal Agents

When the Same Reward Teaches Different Visual Skills

When the Same Reward Teaches Different Visual Skills

Mem++: Non-Destructive Long-Term Memory for Organizational Agents

Mem++: Non-Destructive Long-Term Memory for Organizational Agents

Global Coherence in Multi-Agent Collaboration

Global Coherence in Multi-Agent Collaboration

Sphere Encoder 2: Image Generation with Rotation in a Spherical Latent Space

Sphere Encoder 2: Image Generation with Rotation in a Spherical Latent Space

UTF-8/UTF-16 Routing to Reduce Cross-Script Token Gaps

UTF-8/UTF-16 Routing to Reduce Cross-Script Token Gaps

Interactive Code Execution for Reducing Model Calls in Robot Agents

Interactive Code Execution for Reducing Model Calls in Robot Agents

Turning 2D Vision-Language Models into 3D Programmers with Iterative Feedback

Turning 2D Vision-Language Models into 3D Programmers with Iterative Feedback

ایده‌ها و تجربه‌هایی برای ساختن بهتر.

این هفته را با یک سؤال شروع کن: چه چیزی واقعاً باید ساخته شود؟

مدیریت قبل از ابزار، از وضوح شروع می‌شود.

هوش مصنوعی وقتی ارزش دارد که وارد کار واقعی شود.

فروش خوب، بیشتر از حرف‌زدن، از فهمیدن شروع می‌شود.

این روزها چه چیزی در حال شکل‌گرفتن است؟

چند خط برای مکث، دیدن و دوباره فکرکردن.

چه چیزی از این هفته ماند و چه چیزی باید رها شود؟