Benchmark result · self-reported

LongMemEval-V2, in the sub-second regime.

We measured our memory system on the public LongMemEval-V2 Small benchmark using its own official scorer. Here is the result and everything needed to verify it — published on our terms, whether or not it is ever posted elsewhere.

Self-reported · official scorer · independent verification pending
56.76%
accuracy
451-question Small tier, all questions scored
0.187s
avg. query latency
deterministic memory routing
+3.58
LAFS gain
over the released reference frontier

What this says, and what it does not

LongMemEval-V2's headline metric, LAFS, rewards improving the accuracy-vs-latency frontier — being more accurate at a given speed. Our entry improves that frontier at the fast end: best accuracy in the sub-second regime, roughly six points above the same-speed baseline, hundreds of times faster than the high-accuracy agent baselines that take minutes per query.

This is a self-reported result, measured with the benchmark's official public scorer on the complete Small tier. It is not an accepted leaderboard entry, and it is not the highest raw accuracy overall — the agent baselines reach higher accuracy at 100–180 seconds per query. Our claim is specific: best accuracy-per-latency at the fast end, published with verifiable hashes, pending independent review.

The frontier

SystemAccuracyLatencyNote
RAG (query → slice + notes)51.0%0.2 sreleased baseline
UDI v0.3.1 (ours)56.76%0.187 ssub-second frontier
AgentRunbook-R58.6%26.9 sreleased baseline
Codex (coding agent)69.9%177.2 sreleased baseline
AgentRunbook-C74.9%108.3 sreleased baseline

At the same speed as the fastest baseline, we answer ~6 points more correctly. The high-accuracy systems win on raw accuracy but cost 100–950× the latency.

Verify it yourself

The frozen result package is content-hashed. When the package is reviewed, these digests let anyone confirm the bytes are exactly what produced the number above — nothing swapped, nothing tuned after the fact.

Frozen package · sha256
558307411338f8a7831d2aef…
Submission overview · sha256
259467c906631b7f1b5b8c90…
Code artifact · sha256
230ccf4dbb6d9125f1f6331c…
Protected package · sha256
3a52886c6d30b6ca51123947…

Come falsify us.

The full method, matched baselines, and negative results live in the Lab. If the number is wrong, the fastest way to show it is to reproduce the scorer on the public benchmark — and we will publish what you find, positive or negative.