run_2026-09-26_cohere-rerank
v3.4 · Cohere embed-v4 + rerank-3.5
Claude Sonnet 5.5 · Cohere embed-v4 · dense + rerank-3.5 · support-prompt v7 on support-golden-v4 v4.6
Pass rate
86.1%
Faithfulness
0.91
Answer correctness
0.87
Context precision
0.83
Context recall
0.85
Latency p95
4.3s
Failures by category
172 failed samples · largest share: unsupported
| Category | Share | vs reference |
|---|---|---|
| Unsupported claims | 27.9% | +0.7 pts |
| Incorrect factual answers | 22.1% | +1.1 pts |
| Missing context | 16.9% | −2.3 pts |
| Retrieval mismatch | 15.1% | −2.8 pts |
| Instruction-following failures | 12.8% | +1.7 pts |
| Other | 5.2% | +1.5 pts |
Run details
- Dataset
- support-golden-v4 v4.6 · 1,240 items
- Generator
- Claude Sonnet 5.5
- Retriever
- Cohere embed-v4 · dense + rerank-3.5
- Prompt
- support-prompt v7
- Judge
- LLM judge · Claude Opus 5.5 · rubric v3
- Started
- Sep 26, 09:40 UTC
- Duration
- 3h 2m
- Triggered by
- Sara Lindqvist
- Tokens / request
- 3,410 in · 212 out
- Compared with
- support-assistant v3.4 · release candidate
- retriever
- rerank
All metrics
Against support-assistant v3.4 · release candidate
| Metric | Group | This run | Reference | Change |
|---|---|---|---|---|
| Pass rateShare of samples the judge marked as passing all rubric checks. | Outcome | 86.1% | 86.9% | −0.8 pts |
| FaithfulnessShare of answer claims supported by the retrieved context. | Generation | 0.91 | 0.91 | 0.00 |
| Answer correctnessAgreement with the gold answer on facts, not style. | Generation | 0.87 | 0.88 | −0.01 |
| Instruction adherenceRespect for tone, format, length, language and policy constraints. | Generation | 0.93 | 0.93 | 0.00 |
| Context precisionRank-weighted precision of relevant chunks passed to the generator. | Retrieval | 0.83 | 0.79 | +0.04 |
| Context recallShare of gold-answer statements supported by the retrieved context. | Retrieval | 0.85 | 0.84 | +0.01 |
| Recall@5Share of gold passages found in the top 5 retrieved chunks. | Retrieval | 0.90 | 0.88 | +0.02 |
| MRRMean reciprocal rank of the first relevant chunk. | Retrieval | 0.79 | 0.74 | +0.05 |
| NDCG@10Graded ranking quality of the top 10, normalized by the ideal order. | Retrieval | 0.84 | 0.81 | +0.03 |
| Latency p50Median end-to-end time to final token. | Operational | 2.2s | 1.9s | +0.3s |
| Latency p9595th percentile end-to-end time to final token. | Operational | 4.3s | 3.8s | +0.5s |
| Cost per requestModel, embedding and reranking cost per answered request (USD). | Operational | $0.0049 | $0.0042 | +0.0007 |
Nightly trend
Nightly runs on the support-golden-v4 slice, Jul 11 to Oct 8. Dashed line marks this run.
- Pass rate · nightly
- This run · 86.1%
Across datasets
v3.4 · Cohere rerank on every dataset in the project. Rows marked “matrix” come from the comparison matrix rather than a dedicated run.
| Dataset | Items | Pass rate | Faith. | Ctx recall | Recall@5 | p95 | Source |
|---|---|---|---|---|---|---|---|
| support-golden-v4this run | 1,240 | 86.1% | 0.91 | 0.85 | 0.90 | 4.3s | This run |
| billing-edge-cases | 312 | 81.1% | 0.87 | 0.83 | 0.88 | 4.5s | matrix |
| roaming-and-travel | 188 | 80.3% | 0.88 | 0.80 | 0.85 | 4.4s | matrix |
| device-troubleshooting | 426 | 83.8% | 0.89 | 0.84 | 0.89 | 5.0s | matrix |
| account-security | 154 | 87.0% | 0.92 | 0.86 | 0.91 | 4.2s | matrix |
| locale-formatting | 96 | 79.2% | 0.86 | 0.81 | 0.86 | 4.6s | matrix |
| synthetic-hard-negatives | 540 | 75.7% | 0.86 | 0.74 | 0.79 | 4.5s | matrix |
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.
How long does it take to port my number to Harbor? Number porting Passjudge 0.93 |
Does the Fall 2026 upgrade offer work with a family plan discount? Promotions Passjudge 0.92 |
Can I return a phone I bought online if I opened the box? Returns & warranty Passjudge 0.95 |
My bill says 41,20 €. Is that forty-one euros or four thousand? Billing & payments Passjudge 0.90 |
My eSIM QR code says it has already been used. What should I do? eSIM & activation Passjudge 0.93 |
Is SIM swap protection turned on by default? Account & security Passjudge 0.94 |
If I upgrade my plan mid-month, how is my next bill calculated? Plans & upgrades Passjudge 0.89 |
How much data can I use in the EU on the Unlimited Plus plan? Roaming & travel Passjudge 0.98 |
Why was I charged twice for my September bill? Billing & payments Passjudge 0.87 |
What is the excess on a theft insurance claim? Returns & warranty Passjudge 0.88 |