Forensia CTI Core

Summary

Ingested excerpt (first ~500 chars of normalized text).

Inside VAKRA: Reasoning, Tool Use, and Failure Modes of Agents Inside VAKRA: Reasoning, Tool Use, and Failure Modes of Agents Hugging Face Models Datasets Spaces Buckets new Docs Enterprise Pricing Log In Sign Up Back to Articles Inside VAKRA: Reasoning, Tool Use, and Failure Modes of Agents Enterprise Article Published April 15, 2026 Upvote 28 +22 Ankita Naik ankita-naik Follow ibm-research danish danish Follow ibm-research Ben Ben871 Follow ibm-research Anupama Murthi anup…

https://huggingface.co/blog/ibm-research/vakra-benchmark-analysis

sha256:997806620590f07bdb3a4f9e98712753aff533ac28184b16b0de9da625a267e6

What we pulled out

Deterministic extractor (IOC + allowlisted tokens + ATT&CK IDs present in DB).

cve0
md50
url3
ipv40
ipv60
sha10
domain2
sha2560
malware_token0
attack_pattern0
threat_actor_token0

Indicators

Linked with report → mentions → indicator. Values open the indicator workspace.

Type	Value	Link
domain	github.com	Open →
domain	huggingface.co	Open →
url	https://github.com/IBM/vakra?tab=readme-ov-file#submitting-to-the-live-leaderboard	Open →
url	https://huggingface.co/datasets/ibm-research/VAKRA	Open →
url	https://github.com/IBM/vakra	Open →

Malware families

Allowlist token matches only.

None extracted.

Threat actors mentioned

Allowlist mentions — not a formal attribution verdict.

None extracted.

ATT&CK techniques

MITRE IDs referenced in text and present in local technique table.

None linked (import ATT&CK or adjust text).

CONTINUE INVESTIGATION

High-signal pivots without leaving the thread you started in search.

All reports
Browse the report corpus.
Intel graph (first IOC)
Neighborhood from the first linked indicator.

← Back to reports