aiqbench analysis
This release introduces VectraYX-Vision-1B, a sub-2B parameter model designed specifically for Spanish and LATAM cybersecurity tasks with integrated visual reasoning and native tool use capabilities. The authors have addressed critical benchmarking integrity issues by retracting inflated scores from the initial version and correcting implementation bugs related to LoRA learning rates and checkpoint loading. For researchers in low-resource language security, this represents a significant step toward robust, region-specific AI agents that can interpret visual data and execute tools within constrained environments.
VectraYX-Vision-1B: A Sub-2B Spanish/LATAM Cybersecurity Vision-Language Model with Structured Visual Reasoning and Native Tool Use
arXiv:2608.08477v2 Announce Type: replace Abstract: 21 pages, 1 figure, 9 tables. v2: retracted the B6 tool-id score of 0.08 (v1) after finding 3 benchmark harness bugs; under the fixed harness every B6/B7 metric is 0.0. Fixed a LoRA LR bug and a checkpoint-load-order bug. Added a 9-field gate + linear probe. Reframed the NoPE ablation as open, confound quantified. Code/checkpoints on HF.
Abstract:21 pages, 1 figure, 9 tables. v2: retracted the B6 tool-id score of 0.08 (v1) after finding 3 benchmark harness bugs; under the fixed harness every B6/B7 metric is 0.0. Fixed a LoRA LR bug and a checkpoint-load-order bug. Added a 9-field gate + linear probe. Reframed the NoPE ablation as open, confound quantified. Code/checkpoints on HF.
20 pages, 1 figure, 9 tables. v2: retracted the B6 tool-id score of 0.08 (v1) after finding 3 benchmark harness bugs; under the fixed harness every B6/B7 metric is 0.0. Fixed a LoRA LR bug and a checkpoint-load-order bug. Added a 9-field gate + linear probe. Reframed the NoPE ablation as open, confound quantified. Code/checkpoints on HF
Computation and Language (cs.CL)
arXiv:2608.08477 [cs.CL]
(or arXiv:2608.08477v2 [cs.CL] for this version)
Full text under the publisher's open licence · 126 words
Related coverage

NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI
NVIDIA · May 21, 2026
[AINews] High Taste Testers Ep. 1
Latent Space · Oct 17, 2026
Links #5: 2026/07
LessWrong · Jul 31, 2026
[AINews] AI Engineer NYC: Gemini 4 Argon, Vals AI Benchmarks
Latent Space · Oct 14, 2026