framework

Singapore's IMDA has launched the Model AI Governance Framework for Agentic AI, a voluntary guide addressing key dimensions like access control, accountability, logging, and user transparency to prevent risks in AI agent systems.

LLM-as-a-Verifier is a general-purpose verification framework that provides fine-grained feedback for AI agents, achieving state-of-the-art performance on benchmarks like Terminal-Bench and SWE-Bench.

DE-Venus is a unified framework for data-efficient Reinforcement Learning with Verifiable Rewards in large language models, reducing annotation and training costs while maintaining performance.

TorchSpec is a PyTorch-native framework for training speculative decoding draft models, released in collaboration with vllm and demonstrated with Kimi K3 draft models on NVIDIA GB200 hardware.

SALA is a Semantic-Aware Logical Alignment framework that improves demonstration selection for complex reasoning in in-context learning by automatically learning task-specific reasoning operations and using dynamic time warping for flexible alignment, outperforming existing methods.

FUSE is a unified framework for evaluating dangerous capabilities of large language models, assessing knowledge, defense, and harm dimensions. The study finds that newer models have increased dangerous capabilities despite alignment progress, and it provides cross-model comparisons.

This paper proposes AI Morbidity and Mortality (AI M&M), a blameless framework for case-based review of clinical AI failures, aiming to convert individual errors into actionable institutional learning.

PaperCompiler is a framework that compiles research papers into structured repository specifications to improve code fidelity and reduce errors in paper-to-code generation.

This paper introduces EvoUndo, a framework that enforces recoverability as a constraint in self-evolving AI agents, highlighting bottlenecks like state grounding and recovery-language expressivity.

Teknium announced the release of Hermes Agent v0.21.0 with new features like Bots Mode and Agent 2 Agent Comms, prompting a user to switch from Grokbot.

ORDDAR is a reasoning framework that models cognitive state transitions to detect and repair localized distortions, enhancing AI reasoning quality and interpretability across benchmarks.

WikiSkill is a framework that co-evolves agent skills with a persistent knowledge base, consistently outperforming state-of-the-art methods and enabling effective skill transfer across models.

This article discusses the 4E framework for architectural work, focusing on the 'Evaluate' and 'Examine' steps to identify trade-offs and understand the problem holistically.

Framework has officially announced a new motherboard supporting 192GB memory, with speculation on pricing, PCIe slot availability, and power delivery capabilities.

Kent C. Dodds argues that defining and maintaining well-designed primitives in codebases allows AI agents to operate more autonomously with looser, higher-level specifications, improving effectiveness in software development.

Google for Education shares a 5-step framework by Learning Coach Melissa for using Gemini to generate tailored classroom materials without additional tweaking.

The paper presents a six-stage audit framework for assessing reproducibility in neuro-symbolic AI literature, finding only 6.5% of studies with published artifacts can be reproduced, highlighting a crisis in research reproducibility.

Anthropic has unveiled a new physical AI framework that will soon be open-sourced.

This paper introduces PRISK, a framework for evaluating risks in LLM personalization, finding that personalized context increases irrelevant personalization, preference narrowing, and sycophantic bias across 13 models.

WikiSkill introduces a framework that co-evolves agent skills with a persistent knowledge base to systematically accumulate experience and improve performance across models, demonstrating benefits over state-of-the-art skill-evolution methods.