{"total":17,"items":[{"citing_arxiv_id":"2607.07109","ref_index":25,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Certifying Ghosts: How Cybersecurity AI Agents Break the EU Cyber Resilience Act","primary_cat":"cs.CR","submitted_at":"2026-07-08T07:50:49+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Cybersecurity AI agents invalidate the CRA's premises about vulnerability discovery, exploitation, and remediation speed, making static conformity certificates false and requiring continuous agent-operated defense.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00913","ref_index":62,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Two AI Metrics Diverged: Will it Make All the Difference?","primary_cat":"cs.AI","submitted_at":"2026-07-01T13:18:21+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Bounded performance metrics always favor convergence of AI capabilities to meek models while unbounded metrics allow frontier models to maintain leads indefinitely, with policy implications for capability concentration.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29981","ref_index":35,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Hephaestus: Toward a Cybersecurity AI Scientist","primary_cat":"cs.CR","submitted_at":"2026-06-29T08:56:37+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"The paper proposes the Cybersecurity AI Scientist as a modular multi-agent architecture for automating cybersecurity research, distinguished by its focus on non-stationary threats and anchored in a four-zeros risk-trust-incident-energy frame.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.26216","ref_index":25,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"CyberChainBench: Can AI Agents Secure Smart Contracts Against Real-World On-Chain Vulnerabilities?","primary_cat":"cs.CR","submitted_at":"2026-06-24T17:58:16+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"CyberChainBench is a new benchmark evaluating LLM agents on vulnerability detection, exploit generation, and patch synthesis using 541 real-world DeFi incidents with on-chain historical evaluation, showing peak performance of 43.7% on exploitation.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.14295","ref_index":43,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"AgentCyberRange: Benchmarking Frontier AI Systems in Realistic Cyber Ranges","primary_cat":"cs.CR","submitted_at":"2026-06-12T09:29:58+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"Introduces the first open multi-host cyber range benchmark AgentCyberRange with Cage toolchain and evaluates six frontier AI systems on web exploitation and post-exploitation tasks across 110 vulnerabilities.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.13608","ref_index":46,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"AgentBeats: Agentifying Agent Assessment for Openness, Standardization, and Reproducibility","primary_cat":"cs.AI","submitted_at":"2026-06-11T17:23:54+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"AgentBeats implements agentified evaluation of diverse AI agents through standardized interfaces, validated at scale in a five-month competition with 298 judges and 467 subjects plus a coding case study.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.10281","ref_index":60,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Benchmarking and Exploring the Capabilities of LLMs for Attack Investigations","primary_cat":"cs.CR","submitted_at":"2026-06-09T01:09:04+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"AuditBench is a new benchmark of audit logs from 50+ malicious and benign scenarios that evaluates five LLMs on four security investigation tasks and analyzes their performance and error profiles.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.04460","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"CyberGym-E2E: Scalable Real-World Benchmark for AI Agents' End-to-End Cybersecurity Capabilities","primary_cat":"cs.CR","submitted_at":"2026-06-03T05:06:37+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"CyberGym-E2E turns 920 real OSS-Fuzz vulnerabilities into end-to-end agent tasks; results show frontier agents patch known bugs well (65.9% S3) but find the exact ground-truth vulnerability less often (22.2% S4).","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.26195","ref_index":65,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"CyberEvolver: Structured Self-Evolution for Cybersecurity Agents On the Fly","primary_cat":"cs.CR","submitted_at":"2026-05-25T16:26:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"CyberEvolver introduces a four-layer self-evolving agent architecture with trace-to-diagnosis and population beam search that raises seed agent success rates by 13.6% on CTF, exploitation, and penetration tasks across four LLMs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.12233","ref_index":29,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"No More, No Less: Task Alignment in Terminal Agents","primary_cat":"cs.LG","submitted_at":"2026-05-12T15:06:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"The TAB benchmark reveals that frontier terminal agents achieve high task completion but low selective alignment with relevant environmental cues over distractors, and prompt-injection defenses block both.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.10597","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"CrackMeBench: Binary Reverse Engineering for Agents","primary_cat":"cs.SE","submitted_at":"2026-05-11T14:01:36+00:00","verdict":"ACCEPT","verdict_confidence":"MODERATE","novelty_score":7.0,"formal_verification":"none","one_line_summary":"CrackMeBench introduces 20 deterministic binary validation tasks and reports GPT-5.5 solving 11/12 generated ones at pass@3 while Claude and Kimi lag, especially on harder tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.05531","ref_index":34,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Beyond Collection: Measuring the Detection Efficacy of Modern Security Logging Standards","primary_cat":"cs.CR","submitted_at":"2026-05-07T00:19:31+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"SETC framework provides the first systematic comparison of CIM, OCSF, and ECS logging standards by running 50 RCE exploits and measuring how well each captures attack indicators.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.06713","ref_index":15,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Agentic AI and the Industrialization of Cyber Offense: Forecast, Consequences, and Defensive Priorities for Enterprises and the Mittelstand","primary_cat":"cs.CR","submitted_at":"2026-05-06T23:23:26+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Agentic AI lowers the cost and speed of cyber attacks, requiring immediate improvements in identity management, phishing-resistant authentication, patching, and agent governance for large enterprises and the Mittelstand.","context_count":1,"top_context_role":"background","top_context_polarity":"support","context_text":"It is a local privilege-escalation flaw: not remotely exploitable by itself, but highly consequential once an attacker has any local code execution, for example through compromised cre- dentials, a malicious CI job, or a container foothold. Microsoft emphasizes its impact in cloud, CI/CD, and Kubernetes envi- ronments [14]; Ubuntu rated it high severity and published mitigations and fixes [15]; NVD records it in CISA's Known Exploited Vulnerabilities catalog with a May 15, 2026 reme- diation due date [17]. The lesson is broader than one CVE: when initial access becomes cheaper and privilege escalation becomes easier to operationalize, defenders must assume that the time from foothold to impact will continue to shrink. A. Contributions This paper makes four contributions."},{"citing_arxiv_id":"2604.11950","ref_index":59,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"AnyPoC: Universal Proof-of-Concept Test Generation for Scalable LLM-Based Bug Detection","primary_cat":"cs.SE","submitted_at":"2026-04-13T18:44:02+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"AnyPoC validates LLM bug reports by synthesizing, executing, and independently re-checking proof-of-concept tests, finding 121 new bugs with 108 confirmed across large systems.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.07624","ref_index":40,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Program Analysis Guided LLM Agent for Proof-of-Concept Generation","primary_cat":"cs.SE","submitted_at":"2026-04-08T21:55:18+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"PAGENT integrates static and dynamic program analysis guidance with an LLM agent to improve automated proof-of-concept generation success by 132% over prior agentic methods.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.15763","ref_index":48,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"GLM-5: from Vibe Coding to Agentic Engineering","primary_cat":"cs.LG","submitted_at":"2026-02-17T17:50:56+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"GLM-5 is a foundation model that claims state-of-the-art results on coding benchmarks and superior performance on end-to-end software engineering tasks via new asynchronous RL methods and cost-saving DSA.","context_count":1,"top_context_role":"dataset","top_context_polarity":"use_dataset","context_text":"668.0 Tool-Decathlon 39.2 23.8 35.2 27.8 43.5 36.446.3 Vending-Bench 2 $4,432 $2,377 $1,034 $1,198 $4,967 $5,478$3,591 GDPval-AA Elo 1,409 1,198 1,195 1,288 1,400 1,2011,462 6.1.2 Evaluation of Coding Benchmarks For coding benchmarks, we evaluate LLMs on SWE-bench Verified [ 19], SWE-bench Multilin- gual [53], Terminal Bench 2.0 [45], and CyberGym [48]. For SWE-bench Verified & Multilingual, we use the OpenHands framework using a tailored instruction prompt for GLM-5. For Terminal- Bench 2.0, two agent frameworks (i.e., Terminus-2 and Claude Code) are used, and we also report the performance on a verified Terminal-Bench 2.0 that resolves some ambiguous instructions8. The CyberGym benchmark is evaluated in Claude Code 2."},{"citing_arxiv_id":"2602.02276","ref_index":69,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Kimi K2.5: Visual Agentic Intelligence","primary_cat":"cs.CL","submitted_at":"2026-02-02T16:17:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Kimi K2.5 combines joint text-vision training with an Agent Swarm parallel orchestration framework to reach claimed state-of-the-art results on coding, vision, reasoning, and agent tasks while cutting latency up to 4.5 times.","context_count":1,"top_context_role":"dataset","top_context_polarity":"use_dataset","context_text":"AnswerBench [37], GPQA-Diamond [47], MMLU-Pro [64], SimpleQA Verified [22], AdvancedIF [23], and LongBench v2 [9]. 10 Kimi K2.5TECHNICALREPORT •Coding: SWE-Bench Verified [29], SWE-Bench Pro (public) [16], SWE-Bench Multilingual [29], Terminal Bench 2.0 [39], PaperBench (CodeDev) [52], CyberGym [66], SciCode [56], OJBench (cpp) [65], and Live- CodeBench (v6) [28]. •Agentic Capabilities: BrowseComp [68], WideSearch [69],DeepSearchQA [60], FinSearchComp (T2&T3) [26], Seal-0 [45], GDPVal [43]. •Image Understanding:(math & reasoning)MMMU-Pro [75], MMMU (val) [76], CharXiv (RQ) [67], Math- Vision [61] and MathVista (mini) [36];(vision knowledge)SimpleVQA [13] and WorldVQA 2;(perception) ZeroBench (w/ and w/o tools) [48], BabyVision [12], BLINK [18] and MMVP [57];(OCR & document)OCR-"}],"limit":50,"offset":0}