Round 17 breakdown

disqualifiedchallengerSubmission

Round score

—

degenerate output (hf-007: rep 2: distinct 16-gram ratio 0.086 below harness floor over 10068 chars)

Eligible requests

—

trusted baseline exclusions removed

Failed eligible requests

—

each request counted once

Raw weighted speedup

—

before failure cap and deduction

Workload

Temperature range

0.1 to 1.01 (per prompt)

Request concurrency

Tier scheduling

8k, then 16k; one tier at a time

Output policy

Natural EOS · Minimum 90% of baseline tokens per request

Thinking

Disabled

Context limit

262,144 tokens

The same eligible inputs and output ceilings apply to every engine. Candidates must emit at least 90% of baseline tokens per request; natural output lengths can affect tier completion time. Slots refill until the group drains and stay occupied through protocol completion. Baseline exclusions and final drain can reduce actual concurrency below the configured cap.

The round score uses weighted completion times for all eligible work in each tier. Per-request speedups below are diagnostics, not score contributions. Only trusted baseline exclusions are removed; candidate failures remain in the eligible count and incur the configured failure penalty.

No prompt scores

degenerate output (hf-007: rep 2: distinct 16-gram ratio 0.086 below harness floor over 10068 chars)

Request trace

14 requests

Request	TTFT	Total	Tokens	Slowest gap	Stream
hf-000	127.08ms	22.366s	5120	43.27ms	1758 gaps for 5120 tokens
hf-001	105.25ms	18.846s	4463	42.12ms	1480 gaps for 4463 tokens
hf-002	127.04ms	13.013s

Observed request concurrency

Occupied client slots include protocol completion and final drain. Mean occupancy is weighted by elapsed time; it can be below the requested cap.

Repetition	Group start	Requested	Effective cap	Peak	Mean
1	0.000s	C4	4	4	3.38
1	41.114s	C4	4	4	3.02
2	0.000s	C4	4	4	3.27
2	42.339s	C4	4	4	2.95
3	0.000s	C4	4	4	3.14
3	42.702s	C4	4	4	2.95

Prompt	Candidate char-16	Baseline char-16	Drop	Limit	Result
hf-000	0.9697	0.9800	0.0103	0.1000	Passed
hf-001	0.9054	0.9523	0.0469	0.1000	Passed
hf-002	0.7534	0.8879	0.1346	0.1000	Failed distinct 16-gram ratio 0.753 is more than 0.10 below baseline 0.888 over 13180 chars
hf-003	0.9132	0.8584	-0.0548	0.1000	Passed
hf-004	0.9073	0.9443	0.0370	0.1000	Passed
hf-005	0.8841	0.9511	0.0670	0.1000	Failed rep 1: longest repeated span ratio 0.327 at or above harness ceiling over 19482 chars
hf-007	0.9000	0.8173	-0.0827	0.1000	Failed rep 2: distinct 16-gram ratio 0.086 below harness floor over 10068 chars
hf-008	0.9687	0.9108	-0.0580	0.1000	Passed
hf-009	0.9500	0.9288	-0.0212	0.1000	Passed
hf-010	0.9783	0.9711	-0.0071	0.1000	Passed
hf-011	0.8252	0.9402	0.1150	0.1000	Failed distinct 16-gram ratio 0.825 is more than 0.10 below baseline 0.940 over 22405 chars
hf-013	0.9580	0.9677	0.0098	0.1000	Passed
hf-014	0.9504	0.9505	0.0002	0.1000	Passed
hf-015	0.8661	0.8941	0.0280	0.1000	Passed

Prompt	Repetition	Temperature	Top p
hf-003	1	0.305979	1
hf-001	1	0.901395	1
hf-005	1	0.615361	1
hf-000	1	0.901755	1
hf-008	1	0.685191	1
hf-010	1	0.922572	1
hf-007	1	0.209852	1
hf-009	1	0.915819	1
hf-011	1	0.595321	1
hf-013	1	0.96432	1
hf-002	1	0.849391	1
hf-004	1	0.425489	1
hf-014	1	0.545799	1
hf-015	1	0.11923	1
hf-000	2	0.901755	1
hf-001	2	0.901395	1
hf-005	2	0.615361	1
hf-003	2	0.305979	1
hf-007	2	0.209852	1
hf-008	2	0.685191	1
hf-010	2	0.922572	1
hf-002	2	0.849391	1
hf-009	2	0.915819	1
hf-013	2	0.96432	1
hf-011	2	0.595321	1
hf-004	2	0.425489	1
hf-014	2	0.545799	1
hf-015	2	0.11923	1
hf-001	3	0.901395	1
hf-000	3	0.901755	1
hf-005	3	0.615361	1
hf-003	3	0.305979	1
hf-007	3	0.209852	1
hf-010	3	0.922572	1
hf-008	3	0.685191	1
hf-009	3	0.915819	1
hf-002	3	0.849391	1
hf-013	3	0.96432	1
hf-011	3	0.595321	1
hf-004	3	0.425489	1
hf-014	3	0.545799	1
hf-015	3	0.11923	1

Round 17 breakdown

Workload

Request trace

Observed request concurrency

Round 17 breakdown

Round 17 breakdown

Workload

Request trace

Observed request concurrency

Per-prompt repetition checks

Replay sampling

Correctness

SLA replay

Engine