AI科学家能写论文但难保真，这份综述直击验证缺口的要害。
arXiv:2608.05179v1 Announce Type: cross Abstract: Large language model (LLM) agents are increasingly used across the scientific research lifecycle: id…
语言模型做决策不可靠？YUKTI提出从自然语言到鲁棒可验证决策的新框架，挑战传统单目标优化的置信度陷阱。
arXiv:2607.09706v1 Announce Type: new Abstract: Language models turn a worded situation into a numeric plan, and the dominant pipelines (NL4Opt, OptiM…
用密码学证书为AI行为提供可信证明，在形式验证与加密认证间找到新平衡点。
arXiv:2606.23768v1 Announce Type: cross Abstract: We propose cryptographic certificates of validity for agentic AI systems. The core idea is to formal…
用LLM自动搜索并验证系统启发式策略，实现实例定制化，突破传统手工调优瓶颈。
arXiv:2512.25065v2 Announce Type: replace-cross Abstract: Systems resource management tasks rely primarily on hand-designed heuristics. However, growi…
论文提出一种可验证的反伪造防火墙，让LLM代理在无人监督时无法虚报成功，将诚实作为自主任务的首要度量。
arXiv:2606.11688v1 Announce Type: cross Abstract: Long-horizon LLM agents are not trusted to run unattended: with no human watching, they confidently …
让AI代理行为拥有可验证收据，AERF通过Ed25519签名确保可信溯源。
Article URL: https://github.com/aerf-spec/aerf Comments URL: https://news.ycombinator.com/item?id=48369900 Points: 1 # Comments: 0
大模型数学推理新突破：HERMES实现高效与可验证的推理过程
arXiv:2511.18760v2 Announce Type: replace Abstract: Informal mathematics has been central to modern large language model (LLM) reasoning, offering fle…
提出可验证过程奖励机制，让智能体推理更可信可解释，强化学习新思路。
arXiv:2605.10325v2 Announce Type: replace Abstract: Reinforcement learning from verifiable rewards (RLVR) has improved the reasoning abilities of larg…
专为评估训练大模型设计的可扩展可验证规划基准，填补LLM在复杂规划任务中的性能缺口。
arXiv:2605.20873v1 Announce Type: cross Abstract: Planning is a fundamental capability for large language models (LLMs) because such complex tasks req…
一项可复现、可验证的评估框架，让工具调用型LLM智能体的性能对比不再模糊。
arXiv:2605.15104v2 Announce Type: replace Abstract: Voice agents increasingly require reliable tool use from speech, whereas prominent tool-calling be…
提出可验证的LLM上下文压缩形式化框架，压缩同时保持承诺完整性，AI安全新思路。
arXiv:2605.17304v1 Announce Type: new Abstract: LLM context is not just tokens; it is a set of commitments. Long-running conversations accumulate goal…
58位作者、30家机构联合发布报告，提出10种机制来增强AI系统安全、公平、隐私方面的可验证性。
We’ve contributed to a multi-stakeholder report by 58 co-authors at 30 organizations, including the Centre for the Future of Intelligence, Mila, Schwa…