Labs

Measured. With the sample size on every row.

Measured benchmark results for qbrin against traditional RAG, mem0, Graphiti and HippoRAG 2, with the sample size and the caveat on every row.

Read the agent authorization proof

Measured: we ran it
Published: a competitor’s own number

Headline results

A checked answer, from only the passages it needs.

qbrin gives each question only the passages it needs, then checks the answer before it replies.

$0.0006

per answer, verification call included

Half pays for the answer, half for the check. 12 questions, single run, Aug 20 2026; not the production configuration.

0/500

made-up answers on fake-name trap questions

LlamaIndex’s default pipeline made up 11 (same corpus, same model, hand-verified, Jul 18-19 2026).

93.7%

of answers correct (491 of 524)

On 700 held-out HotpotQA questions qbrin answered 74.9% and declined the others. Single run, Jul 18 2026.

Cost: qbrin bench-cost run, Aug 20 2026. 12 questions (8 factual, 4 with no answer in the data) over a 31K-token knowledge base of qbrin’s own documents; 6 retrieved passages; DeepSeek v4-flash answer model. Single run; not the production configuration. The trap and accuracy figures come from a separate audited run (Jul 18-19 2026) with a different configuration; see the HotpotQA suite below.

Proof, measured

Built not to make things up.

Numbers come from real benchmark runs on qbrin’s own pipeline: the ERB set of 500 real company questions (470 with a known correct source), plus trap questions built to bait a wrong answer. Each figure is labelled measured (we ran it) or published (a competitor’s own number).

Made-up answers

Across 120 trap questions on four corpora — invented projects, false dates, wrong values — qbrin invented zero: it declines, or corrects the false premise with the cited real fact. (Measured 2026-07: 0 invented / 120 traps, groundedness-audited.)

86%

Citations you can trust

With its double-checking layer on, a cited source genuinely backs the answer 86% of the time, raw keyword search manages just 6%. (Measured, ERB.)

100%

The current fact, not the stale one

When a fact changes, qbrin returns the new value every time; mem0 (open-source) keeps the stale one 3 times in 4. (Measured, n=8.)

88%

Finds the right source

Retrieval lands the correct source document in the candidate set 88% of the time, and the double-checking layer then keeps the cited ones honest. (Measured, ERB.)

No per-suite second-by-second timings are shown unless we measured them. The full SOC write-up, qbrin against a Wazuh RAG pipeline, an MCP agent and a naive LLM analyst, is on the AI in the SOC page, and the architecture-level differences against Glean, including where Glean is stronger, are set out in qbrin vs Glean.

Measured on real 10-K filings

Grounded in the source, not guessing.

Ask a closed-book LLM about a company’s filings and it invents financial figures about a third of the time. qbrin reads the actual document and answers from it, so it’s right, or it tells you it isn’t sure. The same cited-or-abstain contract is what the REST API returns.

Closed-book LLM vs qbrin’s retrieval-grounded answers on FinanceBench, real public-company 10-K filings, each graded against the source passage. Preliminary sample (n=16).

Made-up figures (confidently wrong) · lower is better

Closed-book LLM31%

qbrin (grounded)0%

Measured on FinanceBench 10-K filings, each answer graded against the source passage. Preliminary sample, n=16.

Answer accuracy

Closed-book LLM56%

qbrin (grounded)75%

Measured on FinanceBench 10-K filings, each answer graded against the source passage. Preliminary sample, n=16.

Answered honestly, right, or “I’m not sure”

Closed-book LLM69%

qbrin (grounded)100%

Measured on FinanceBench 10-K filings, each answer graded against the source passage. Preliminary sample, n=16.

Compared with named systems

Against named systems, with the method on every card.

Each card shows the number, what it measures and how it was measured. Open “Methodology and caveats” for the sample size and the limits.

vs LlamaIndex

No made-up answers on 500 fake-name traps

0 / 500qbrin · measured

11 / 500LlamaIndex (default pipeline)

made-up answers on fake-name trap questions (lower is better)

0 made-up answers on 500 fake-name trap questions; LlamaIndex’s default pipeline made up 11 (same corpus, same model, hand-verified, Jul 18-19 2026).

Methodology and caveats

MEASURED, Jul 18-19 2026: 500 fake-name trap questions, same corpus and same answer model for both, every flagged response hand-verified. Precision and coverage for the same run are in the HotpotQA suite below.

vs mem0

Returns the new fact, not the stale one

100%qbrin · measured

25%mem0 (open-source)

correct after a fact changes

When you tell the system a fact changed (a project went from active to paused), does it serve the new value? qbrin does every time; mem0 OSS keeps the outdated fact 3 times out of 4.

Methodology and caveats

MEASURED, n=8, same models for both. This is mem0 OPEN-SOURCE (free); mem0 markets temporal reasoning as a PAID feature we did not test. qbrin’s 100% is its CTP-Bench reference.

vs Zep / Graphiti

Reliably retires superseded facts

100%qbrin · measured

~1 / 6Graphiti

clean fact-invalidations

Against the one rival that genuinely tracks when facts become valid or invalid, does it consistently retire old ones? Graphiti got it right once in six scenarios and formed no link at all in half of them.

Methodology and caveats

MEASURED, 6 scenarios, edges read directly from the graph store (Graphiti’s own search path errored). Graphiti’s bi-temporal IS real but inconsistent; better prompting might lift it. qbrin’s 100% is the CTP-Bench reference.

vs HippoRAG 2

Finds chained facts across documents

97.0qbrin · measured

96.0HippoRAG 2

recall@5 (higher is better)

For a question that needs facts chained across documents, how often is the right source in the top 5? On this news corpus qbrin edges out HippoRAG’s graph algorithm.

Methodology and caveats

MEASURED, same 609-doc corpus, 200 questions, same gold. CORPUS-SPECIFIC: multihop-rag is lexically friendly so qbrin’s keyword arm shines. Win is +1.0pt at R@5 and a tie (99.0) at R@10, not a universal retrieval win.

qbrin vs Glean

The architecture-level differences, including where Glean is stronger.

Read the comparison

Enterprise RAG Benchmark

qbrin’s own enterprise benchmark on real company Q&A.

500 real company questions (470 answerable + 30 unanswerable), the most product-representative test we run. The flagship result is safety plus citations you can trust, not a blanket accuracy crown.

0/281

trap questions answered wrong

Every trap question rejected. Measured.

86%

of citations actually hold

A cited source genuinely backs the answer 86% of the time; raw search: 6%. Measured.

87%

decisions correct, held out

About 87%, on 500 real company questions, with full verification. The tuned policy scores 88.4% on the bank it was tuned against; ~87.2% is what transfers to held-out questions, so that is the number we quote. Measured.

Measured coverage

Where we lead. Where we do not.

Across the runs below, qbrin has the lowest rate of confident wrong answers and the highest citation trust among the systems we compared. Its retrieval is competitive-to-leading once tuned to a corpus, with one architectural exception: relational-graph traversal (a rival’s home turf), where it trails. On raw all-answerable public Q&A it runs a calibration posture (abstains rather than guess); the 2026 embedder upgrade lifted that retrieval substantially (FinanceBench gold-in-context +18pp same-harness). Every cell below is a real run.

datasets

Measured

competing systems

Measured

dimensions

Measured

LeadsLeads: qbrin leads on this axis, no competitor beats it
CompetitiveCompetitive: ties or holds its own
CalibrationTrades for calibration: abstains more, trails raw accuracy, by design
not measured on that axis

Which datasets measured which dimensions, and whether qbrin leads, is competitive, or trades raw accuracy for calibration
Dataset	Safetyfewer made-up answers	Citation-trustcited source actually holds	Current-factretires stale / contested	Retrievalfinds the right source	Answer accuracycorrect when it answers	Multilingualparity across languages
CTP-Bench (temporal/contested)n=12	—	—	Leads	—	—	—
Temporal vs mem0 (open-source)n=8	—	—	Leads	—	—	—
Temporal vs Zep / Graphitin=6	—	—	Leads	—	—	—
PrecisionMemBench (memory-precision)n=77	Leads	Leads	—	—	—	—
BrainBench (relational retrieval)n=145	—	—	—	Calibration	—	—
Multi-hop vs HippoRAG 2n=200	—	—	—	Leads	—	—
MuSiQue (hard multi-hop)n=300	—	—	—	Competitive	—	—
multihop-rag (retrieval)n=2255	—	—	—	Leads	—	—
multihop-rag (answer)n=120	—	—	—	—	Calibration	—
ragbenchn=100	—	—	—	Calibration	Competitive	—
financebenchn=150	—	—	—	Leads	Calibration	—
Enterprise RAG Benchmark (ERB)n=500	Leads	Leads	—	Leads	Calibration	—
G3 contested enterprise Q&An=60	—	Leads	—	—	Competitive	—
Multilingual (en / hi / te / ta)n=188	—	—	—	Leads	Competitive	Leads
Knowledge-map compression (SOTA)n=80	—	—	—	—	Competitive	—
Hallucination / fake-entity setn=60	Leads	—	—	—	—	—
Broad RAG general Q&An=120	Competitive	—	—	—	Competitive	—
Internal enterprise benchmarkn=40	Competitive	Competitive	—	—	Competitive	—
False-premise safety (real / fake)n=55	Competitive	—	—	—	—	—

Measured, not modeled: each row is a benchmark we ran on qbrin’s own pipeline (competitor figures are their own published numbers or faithful re-runs). Several wins are corpus-specific and several safety A/Bs are honest ties; the suite tables below carry the exact figures and caveats. On trust and safety under abstention, no competitor we tested beat qbrin. Its retrieval is competitive-to-leading once tuned to a corpus (except relational-graph traversal, where it trails), and on raw answerable-Q&A accuracy it runs a calibration posture (trades coverage for caution).

Cost per answer

Cost and time per answer.

What one answer costs

$0.0006 per answer, verification call included: half pays for the answer, half for the check.

qbrin bench-cost run, Aug 20 2026: 12 questions (8 factual, 4 with no answer in the data) over a 31K-token knowledge base of qbrin’s own documents; 6 retrieved passages; DeepSeek v4-flash answer model. Single run; not the production configuration.

How long does an answer take?

We only show timings we actually measured. On the real pipeline, a typical answer takes ~2–4 seconds, and almost all of that is the model writing the answer, not the search. Finding the source takes a small fraction (~0.15s).

Median time to find the source, write the answer, and in total, by path
Path	Find source	Write answer	Total (typical)
Map only (lightest)	~0.15s	~2.3s	2.3s
Hybrid chunks	~0.15s	~3.1s	3.2s
Source selection	~0.15s	~4.1s	4.2s
Full stack (heaviest)	~0.15s	~4.0s	4.0s

MEASURED, median (p50) per-query timing on the real pipeline, n=60 each. “Total” is the p50; the 90th-percentile tail runs 6.8–8.5s. We deliberately do NOT publish competitor seconds we did not measure.

Safety and traps

Made-up answers on trap questions.

We feed the system questions deliberately built to bait a wrong answer: made-up names, false premises, missing facts. The honest measure is how often it invents an answer instead of saying “I don’t have that”. On the trap sets below, qbrin invented zero.

0/16

Discriminating traps

Confident wrong answers on the discriminating trap set. Measured.

0/30

Fake-entity premises

Confident wrong answers on the fake-entity premises. Measured.

0/281

ERB powered negatives

Confident wrong answers on the ERB trap questions. Measured.

Confident wrong answers on these trap sets: zero (lower is better).

500 fake-name trap questions

0 made-up answers on 500 fake-name trap questions; LlamaIndex’s default pipeline made up 11 (same corpus, same model, hand-verified, Jul 18-19 2026).

Grounded guardrail in action

When the connected files can’t satisfy a request, ordinary AI tends to guess something plausible. qbrin checks that the answer is actually backed by a source before replying, so it tells you what it could and couldn’t find.

Example trap question

“Send the PDF Maya approved in July 2025 regarding the Q3 budget shifts.”

Query (baiting missing information).

Ordinary AI

Makes something up

“Attached is the Q3 Budget PDF showing Maya’s approval…” (actually pulled a June draft from another author)

qbrin (grounded guardrail)

Need more evidence

“I cannot find a Q3 budget PDF approved by Maya in July 2025. I found a June draft by Sam, and a July thread where Maya requested budget edits, but no final approved PDF.”

Catching subtle wrong answers

Beyond outright invention, a deterministic gate (under a second, no extra AI call) catches sneaky errors. Compared to the verifier on its own, the combined gate is dramatically better at spotting drift.

Share of each error type caught by the verifier alone and by the combined gate
Error type	Verifier alone	Combined gate	Plain meaning
Made-up entity	62%	100%	Every fabricated name caught
Citation contradicts source	60%	88%	Most bad citations caught
Off-by-one / date drift	27%	86%	+59 points over verifier alone

MEASURED, 174-row labeled benchmark, deterministic gate. HONEST LIMIT: this is NOT “zero hallucination”, about 9% of unsupported answers still slip through and ~19% of correct answers are over-flagged.

Benchmark suites

Every suite, with its caveats.

Six suites, every row, and the methodology and caveats under each one.

ERB, Enterprise company Q&A (500 questions)

Real company questions. We measure whether the right source is found, whether the final decision is correct, and how often the system confidently gives a WRONG answer (risk). qbrin’s win here is safety: zero confident wrong answers on the trap questions.

ERB, Enterprise company Q&A (500 questions)
System	Decision / answer accuracy	Coverage (answered)	Confident wrong answers	Risk	Notes
qbrin, full verification	~87%	87.0%	0	5.3%	Best verification policy. Zero false-accepts on the trap questions (281/281 rejected). Accuracy is the held-out figure: the policy scores 88.4% on the bank it was tuned against, ~87.2% on held-out questions.
qbrin, high-stakes citation mode	n/a	43%	n/a	precision 86%	High-stakes mode: a cited source backs the answer 86% of the time (raw search: 6%). Buys precision by answering fewer questions.
Naive RAG (always answers)	~83%	100%	many	11.8%	On this 94%-answerable workload its blanket accuracy is high, but it confidently answers wrong more often. qbrin’s edge is the risk/safety axis, not headline accuracy.
Hybrid search, no double-check	n/a	100%	n/a	precision 6%	Retrieval finds the source 88% of the time, but with no verification layer only 6% of citations actually back the claim.

Methodology and caveats

MEASURED, ERB 500-question enterprise benchmark (470 answerable + 30 unanswerable). Honest note: on a 94%-answerable workload, abstaining COSTS net accuracy vs naive RAG, qbrin’s real win is risk/safety, not blanket accuracy. Second honest note: the accuracy column is the HELD-OUT figure. The verification policy was selected against this same question bank, which puts its in-sample score at 88.4%; roughly 87.2% transfers to held-out questions, and that is what we publish. The false-accept and risk columns are not affected by that selection.

HotpotQA held-out, four-system benchmark on real human gold

Our strictest trust-layer test — built to survive hostile due diligence. 700 held-out HotpotQA questions with human-labeled gold (zero overlap with anything used during development), plus 500 unique realistic-name fabricated-entity traps and 150 corrupted-value premises. Four systems, one run each, same day, on the identical corpus, answer model, and embeddings. One scorer for all: shared refusal detector, then string-containment, then an LLM judge for paraphrases. Raw outputs retained for every row; every machine-flagged trap response hand-verified; a stratified sample independently re-judged (92.4% agreement, no directional skew).

HotpotQA held-out, four-system benchmark on real human gold
System	Precision when answering	Coverage (answered)	Made-up-entity hallucination	Notes
qbrin (full verification pipeline)	93.7%	74.9%	0 / 500	Zero fabrications across all 500 traps, hand-verified: 499 explicit refusals plus 1 response that flagged the name mismatch and answered about the real, cited person instead. Abstains rather than guesses.
qbrin (retrieval pinned to baseline k=8)	93.4%	69.4%	0 / 499*	Same retrieval budget as the baselines, consensus off — and the result barely moves. The safety comes from the verification layer, not a bigger search. (*1 API-error row excluded.)
LlamaIndex (default pipeline, top-k 8)	84.7%	96.4%	11 / 500	Hand-verified from raw outputs: 8 direct fabrications about nonexistent people + 3 silent substitutions of a different real entity. 0-vs-11 is statistically significant (one-sided p < 0.001).
Naive RAG (top-k 8, no abstention license)	80.5%	97.3%	≥155 / 500	Its prompt never invites abstention, so treat this as a floor: hand-verified from sampled flagged rows at 31% minimum (best estimate ~38%).

Methodology and caveats

MEASURED 2026-07-18/19, N=1,350 per system (700 held-out HotpotQA distractor-validation questions with human gold + 500 unique realistic-name fabricated-entity traps + 150 corrupted-value premises). Four systems, single run each, same day: identical 6,878-passage corpus, identical answer model, identical bge-m3 embeddings. One scorer for all arms; raw outputs retained for every row; every machine-flagged trap response hand-read; independent-judge sample agreed 92.4% with no directional skew; adversarially audited. EXPLICIT non-claims: qbrin is NOT zero-wrong (93.7% precision = 33 wrong of 524 answered) and does NOT lead on coverage — it abstains on roughly 1 in 4 answerable questions rather than guess. Corrupted-value premise acceptance was 0 for ALL four systems (not a differentiator). The v1 numbers previously shown here (56%→0%, 62→82) are superseded by this audited held-out run.

Multi-hop retrieval, head-to-head vs HippoRAG

Questions that need facts chained across several documents. We report recall@5 (how often the right source is in the top 5) on the same corpus and gold set as HippoRAG 2.

Multi-hop retrieval, head-to-head vs HippoRAG
System	Recall@5	Recall@10	Notes
qbrin, keyword (BM25) arm	97.0	99.0	Edges out HippoRAG on this lexically-friendly news corpus. +1.0pt @5, tie @10.
qbrin, hybrid arm	94.0	98.5	Blends keyword + meaning.
qbrin, meaning (vector) arm	90.0	95.0	Self-hosted embedding (no per-query API cost).
HippoRAG 2 (graph algorithm)	96.0	99.0	Run full-KG, API mode, same 609-doc corpus / 200 questions / same gold.

Methodology and caveats

MEASURED, multihop-rag, 609 docs, 200 questions, identical gold. CORPUS-SPECIFIC: this corpus is lexically friendly, so the keyword arm excels, NOT a universal retrieval win. On the harder MuSiQue set, qbrin’s query-decomposition arm scores 64.4 vs HippoRAG’s published 51.9, but that comparison is NOT matched (different pool).

Knowledge-map compression vs the full prior-art field

When the knowledge map must fit a token budget, whose compression keeps the most answerable facts? This is the full SOTA comparison, qbrin against RAPTOR, GraphRAG, vanilla RAG, and community-summary on the SAME real inbox graph, with the same model across all methods, judge, questions, and budgets. Cells are overall answer-accuracy %. Honest read: qbrin is built to win at large budgets and on decision/relationship questions, it leads at 8192 and clearly beats GraphRAG, but it does NOT sweep every budget.

Knowledge-map compression vs the full prior-art field
Token budget	qbrin	Community-summary	RAPTOR	GraphRAG	Vanilla RAG	Notes
1,024 tokens	10.0	7.5	8.8	1.3	6.3	qbrin leads at the tightest budget. Top three (qbrin/RAPTOR/community) within overlapping 95% CIs; GraphRAG already collapsing.
2,048 tokens	11.3	12.5	11.3	2.5	7.5	Community-summary edges ahead mid-budget; qbrin ≈ RAPTOR. CIs overlap.
4,096 tokens	12.5	22.5	16.3	3.8	12.5	Community-summary best here; RAPTOR 2nd. qbrin not yet at its design budget.
8,192 tokens	33.8	27.5	18.8	5.0	12.5	qbrin LEADS at the largest budget, its design target, ahead of community (27.5) and RAPTOR (18.8). Still overlapping CIs at n=80.
Decision / relationship Qs @8192	30.0	20.0	22.5	2.5	12.5	qbrin’s thesis: on decision/project questions at the large budget it leads clearly (30.0 vs community 20.0, RAPTOR 22.5), consistent with how qbrin is designed.

Methodology and caveats

MEASURED, n=80 (40 hub / 40 evidence), real inbox graph, ALL arms on the same model + judge (uniform, reasoning disabled across arms), bootstrap 95% CIs. HONEST READ, NO method sweeps: qbrin wins at 8192 (33.8%) and on decision/relationship questions (30.0%), and clearly beats GraphRAG at every budget (1–5%, because abstractive community reports lose the specific facts). But community-summary wins the mid-budgets (2048/4096) and RAPTOR is genuinely competitive (2nd–3rd every budget), at most budgets the top three (qbrin/community/RAPTOR) have OVERLAPPING 95% CIs, so n=80 is underpowered to separate them. Positioning: best at large budgets and on decisions, and the clear winner vs GraphRAG, NOT a universal compression champion. Absolute accuracy is low because this is a large real, noisy inbox graph run with reasoning disabled (uniform across arms).

Public benchmarks, answer accuracy

Standard all-answerable public Q&A sets. The 2026 embedder upgrade substantially improved retrieval here, same-harness FinanceBench gold-in-context jumped 44%→62% (and ~9%→62% vs the prior production chunking). qbrin keeps a calibration posture (abstains rather than guess), so on raw all-answerable accuracy it favors precision over coverage. The accuracy columns below were measured before the embedder upgrade.

Public benchmarks, answer accuracy
System	Decision / answer accuracy	Traditional RAG	Notes
qbrin, financebench	20.0%	31.3%	Strict independent-judge baseline (PRE-upgrade embedder). The embedder upgrade lifted same-harness retrieval +18pp (gold-in-top8 44→62%) and accuracy +9pp (35→44%, N=150, self-judge) at flat tokens, with a small +2pp wrong-rate, a genuine gain; the accuracy columns were measured before it.
qbrin, multihop-rag	72.5%	77.5%	Strict independent-judge baseline. The embedder upgrade lifted same-harness multi-hop retrieval to 80% gold-recall@8 (above the prior ~66–71% ceiling); the answer-accuracy columns were measured before it. qbrin runs a calibration posture (abstains rather than guess) on these all-answerable sets.

Methodology and caveats

MEASURED, strict judge-INDEPENDENT baseline (neutral third-party answer model + independent judge), financebench N=150 / multihop-rag N=120, all-answerable gold. The acc/tradacc columns are that PRE-upgrade baseline. The 2026 embedder upgrade (same-harness, N=150) lifted FinanceBench retrieval +18pp and accuracy +9pp at flat tokens, a real gain that these conservative pre-upgrade columns do not reflect. qbrin’s core safety / false-accept edge is UNTESTABLE on these all-gold corpora (0 unanswerable traps); see the ERB flagship for that.

Catching subtle wrong answers, the verifier gate

How well does the deterministic gate catch sneaky errors: a date off by one, a made-up name, or a citation that doesn’t actually support the claim? Higher catch rate is better.

Catching subtle wrong answers, the verifier gate
Error type	qbrin gate catch rate	Verifier alone	Notes
Fabricated entity (made-up name)	100%	62%	Combined gate catches every made-up entity in the test set.
Citation that contradicts the source	88%	60%	Catches most citations that don’t back the claim.
Off-by-one / date drift	86%	27%	Combined gate adds +59pt over the verifier alone.

Methodology and caveats

MEASURED, 174-row labeled benchmark, deterministic (<1s, no LLM). “base” = verifier alone; “catch” = combined gate. EXPLICIT non-claim: this is NOT “zero hallucination”, ~9% of unsupported answers still leak and ~19% of correct answers are wrongly flagged. NOTE: this 86% is a DIFFERENT metric from the ERB citation 86%.

How we measure

How to read these numbers.

Every figure on this page comes from a real benchmark run on qbrin’s own pipeline and is labelled measured (we ran it) or published (a competitor’s own reported number). We never invent a per-suite second or accuracy. Where a comparison isn’t strictly like-for-like, or a sample is small, the footnote says so. Below is the full set of indicators we audit across quality, safety, economics, and root-cause diagnosis.

Quality

Overall accuracyPercentage of all questions answered correctly.
Answered accuracyAccuracy rate restricted only to questions the system chose to answer.
Exact match / task scoreAutomated validation comparing exact text/values against the gold standard.
CoveragePercentage of queries where the system generated an answer rather than abstaining.

Safety

False accept ratePercentage of queries where the system accepted incorrect context or irrelevant documents as true.
Hallucination rateFrequency of inventing facts, numbers, or citing sources that do not exist.
Abstain ratePercentage of trick or out-of-distribution questions where the system correctly chose to state it doesn't know.
Citation validityThe proportion of cited references that genuinely support the claims made in the answer.

Economics

Input tokensVolume of context tokens read per query.
Output tokensVolume of tokens generated in the response.
Total tokensSum of input and output tokens consumed.
Latency p50 / p95Median and 95th-percentile response times in seconds.
Cost per correct answerTotal dollar cost normalized by the number of correct responses generated.

Diagnosis

Retrieval contains gold?Whether the retrieval step succeeded in finding the correct reference document.
Verifier rejected?Whether the verification module intercepted and blocked a bad model output.
Model failed despite gold?LLM reasoning failures despite having the correct context in its input.
Wrong entity / wrong period?Errors caused by extracting the wrong named entity or mixing up target timeframes.
Wrong source?Failures where the wrong tool, database, or document was searched.

Agent authorization proof

A real LangGraph agent, with and without qbrin: five scripted cases, four unsafe actions without a check, none with it.

See the proof

AI in the SOC

A measured security-alert comparison against a RAG pipeline, an MCP agent and a naive LLM analyst.

See the comparison

See it answer your hardest question.

Bring one real question your team keeps re-asking. We’ll connect a source, read-only, and show you the answer, sourced, in seconds, in a 20-minute walkthrough. Nothing changes in your tools.

Book a demo