Skip to content

run_2026-08-07_nightly

Nightly · v3.3 · 2026-08-07

Claude Sonnet 5.5 · bge-m3 · hybrid (BM25 + dense, RRF) · support-prompt v6 on support-golden-v4 v4.6

CompletedNightlyIn productionsupport-assistant v3.3v2 · 512 tokens / 64 overlap
Compare

Stratified slice of support-golden-v4 (412 items).

Pass rate

81.1%

+0.1 pts(improvement)vs previous nightly334 of 412 samples passed

Faithfulness

0.88

0.00(no change)vs previous nightly

Answer correctness

0.84

−0.01(regression)vs previous nightly

Context precision

0.75

+0.01(improvement)vs previous nightly

Context recall

0.80

−0.01(regression)vs previous nightly

Latency p95

3.6s

−0.1s(improvement)vs previous nightly

Failures by category

78 failed samples · largest share: unsupported

Share of failed samples by category, compared with Nightly · v3.3 · 2026-08-06
CategorySharevs reference
Unsupported claims25.6%−0.7 pts
Incorrect factual answers20.5%+0.8 pts
Missing context21.8%+0.7 pts
Retrieval mismatch20.5%−0.5 pts
Instruction-following failures9.0%−0.2 pts
Other2.6%−0.1 pts

Run details

Dataset
support-golden-v4 v4.6 · 412 items
Generator
Claude Sonnet 5.5
Retriever
bge-m3 · hybrid (BM25 + dense, RRF)
Prompt
support-prompt v6
Judge
LLM judge · Claude Opus 5.5 · rubric v3
Started
Aug 7, 02:00 UTC
Duration
45m 47s
Triggered by
Scheduler
Tokens / request
3,980 in · 221 out
  • nightly
  • golden-slice

All metrics

Against Nightly · v3.3 · 2026-08-06

All metrics for Nightly · v3.3 · 2026-08-07
MetricThis runReferenceChange
Pass rate81.1%81.0%+0.1 pts
Faithfulness0.880.880.00
Answer correctness0.840.85−0.01
Instruction adherence0.890.880.00
Context precision0.750.74+0.01
Context recall0.800.80−0.01
Recall@50.830.830.00
MRR0.700.69+0.01
NDCG@100.760.77−0.01
Latency p501.8s1.7s+0.1s
Latency p953.6s3.7s−0.1s
Cost per request$0.0046$0.00460.0000

Nightly trend

Nightly runs on the support-golden-v4 slice, Jul 11 to Oct 8. This run is one point in the series.

Pass rate moved from 78.5% to 87.2% across 90 nightly runs. Configuration changes: v3.3 on Aug 4, v3.4 on Sep 8.

Across datasets

support-assistant v3.3 on every dataset in the project. Rows marked “matrix” come from the comparison matrix rather than a dedicated run.

Configuration results per dataset
DatasetItemsPass rateFaith.Ctx recallRecall@5p95Source
support-golden-v4this run1,240
81.1%
0.880.800.833.6sThis run
billing-edge-cases312
76.0%
0.830.770.813.9sRun · Aug 20
roaming-and-travel188
75.5%
0.840.740.783.8smatrix
device-troubleshooting426
78.9%
0.850.780.824.4smatrix
account-security154
82.5%
0.880.800.843.6smatrix
locale-formatting96
75.0%
0.820.750.794.0smatrix
synthetic-hard-negatives540
70.7%
0.820.680.723.9smatrix

Traced samples

17 samples with full traces (4 failing, 13 passing), stratified by domain. Open a row to inspect retrieval and the judge's reasoning.

Traced samples for Nightly · v3.3 · 2026-08-07

How much data can I use in Germany on Unlimited Plus before I get charged?

Roaming & travel

FailMismatchjudge 0.18

Why was I charged twice this month?

Billing & payments

FailUnsupportedjudge 0.35

Can I keep my old number if I move from prepaid to a contract plan?

Plans & upgrades

FailIncorrectjudge 0.48

Is there a restocking fee if I return an opened phone?

Returns & warranty

FailMissingjudge 0.13

Does the Fall 2026 upgrade offer work with a family plan discount?

Promotions

Passjudge 0.90

What is the late fee if I pay my bill after the due date?

Billing & payments

Passjudge 0.98

How long does it take to port my number to Harbor?

Number porting

Passjudge 0.92

How do I get a copy of the personal data Harbor holds about me?

Data privacy requests

Passjudge 0.91

My phone shows 4G but mobile data does not work after the iOS update.

Device troubleshooting

Passjudge 0.96

Is SIM swap protection turned on by default?

Account & security

Passjudge 0.87