برچسب: مدل‌های زبانی

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 models more reliable on real tasks in Windows.

A Four-Axis Taxonomy for RAG Systems

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

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 peer-review status is unconfirmed.

Language Drift in RLVR Post-Training of Language Models

This preprint examines why RLVR on novel reasoning tasks may move language-model chains of thought away from human-readable language.

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 feedback; this is a preprint with unverified independent review, and runtime cost,…

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 agent on 700 simulated instances; however, generalization to physical robots and other…

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 disparity; language-model quality results remain mostly close and mixed.