chini-train-heldout-0019-dp6-adversarial

Free-Tier Signup Under Throwaway-Account Flood

adversarial adversarial problem: free-tier signup under throwaway-account flood

Source: chini-train synth generator v0.1

Prompt

Design a system for: free-tier signup under throwaway-account flood (domain: adversarial).

Tier DP6 (adversarial). 11-15 nodes, all scenarios at max, adversarial-heavy. Upper-bound probe.

Constraints:
- At most 14 components on the canvas.
- Monthly cost ceiling: $211 USD. Required behaviors: ratelimit, queue, circuitbreaker, retry, batch.

Return a Chinilla CanvasState that handles the listed scenarios. Include trigger components for each entry point and at least one terminal storage / sink so the simulator can score delivery.

Constraints

Max components: 14
Required behaviors: ratelimit, queue, circuitbreaker, retry, batch
Monthly budget: $211

Stress scenarios

Baseline traffic

baseline

Steady ambient load with no failures.

Adversarial burst

adversarial

Hostile packets injected on top of clean traffic. Defenses must block them without dropping good requests.

Dependency outage

outage

A downstream component is disabled. System must degrade gracefully.

Traffic spike

spike

Traffic suddenly multiplies. The hot path must hold.

Cascading failure

cascade

An initial fault propagates through dependent components.

Latency injection

latency

Extra latency injected into a critical component. Tests degradation behavior under slow downstreams.

Pass criteria (overall)

Min stability score: 90
Max drop rate: 4.8%
Min delivery rate: 93.8%
Max errors: 3

Submit your run

Submissions go through the chini-bench CLI. It calls your model with your key, scores the result locally, and posts to the leaderboard. Nothing leaves your machine except the canvas it produces.

End-to-end:

pip install git+https://github.com/collapseindex/chini-bench-cli.git
export OPENROUTER_API_KEY=...

chini-bench run chini-train-heldout-0019-dp6-adversarial \
  --provider openrouter --model google/gemini-2.0-flash-001 \
  --as alice

Or inspect the prompt first:

chini-bench prompt chini-train-heldout-0019-dp6-adversarial

Providers: openai · anthropic · google · openrouter · ollama

Leaderboard

Rank	Submitter	Model	Score	Stability	Delivery	Design	Pass
#1	chini-train-08	fmt_a_v4_opus_7b single-shot	84	57.0	73.0	100.0	✗
#2	chini-train-08	fmt_a_v4_opus_7b single-shot	84	57.0	73.0	100.0	✗
#3	chini-train-08	rl_v07_pilot_a10b_k8_s2 single-shot	79	55.0	46.0	100.0	✗
#4	chini-train-08	rl_v07_pilot_a10b_k8_s3 single-shot	79	56.0	47.0	100.0	✗
#5	chini-train-08	rl_v07_pilot_a10b_k8_s7 single-shot	79	56.0	48.0	100.0	✗
#6	chini-train-08	rl_v07_full single-shot	79	55.0	46.0	100.0	✗
#7	chini-train-08	fmt_a_7b single-shot	78	54.0	45.0	100.0	✗
#8	chini-train-08	fmt_a_7b single-shot	78	54.0	45.0	100.0	✗
#9	chini-train-08	fmt_a_7b single-shot	78	54.0	45.0	100.0	✗
#10	chini-train-08	rl_v06_run2 single-shot	78	54.0	44.0	100.0	✗
#11	chini-train-08	rl_v07_pilot_a10b_k8_s0 single-shot	78	54.0	45.0	100.0	✗
#12	chini-train-08	rl_v07_pilot_a10b_k8_s4 single-shot	78	54.0	44.0	100.0	✗
#13	chini-train-08	rl_v07_pilot_a10b_k8_s5 single-shot	78	53.0	43.0	100.0	✗
#14	chini-train-08	fmt_a_v5_mixed_7b single-shot	77	51.0	41.0	100.0	✗
#15	chini-train-08	fmt_a_v5 single-shot	77	51.0	41.0	100.0	✗
#16	chini-train-08	rl_v07_pilot_a10b_k8_s1 single-shot	77	52.0	42.0	100.0	✗
#17	chini-train-08	rl_v07_pilot_a10b_k8_s6 single-shot	77	52.0	42.0	100.0	✗
#18	chini-train-08	fmtA single-shot	77	51.0	41.0	100.0	✗
#19	chini-train-08	base single-shot	73	56.0	47.0	100.0	✗
#20	chini-train-08	fmt_a single-shot	73	57.0	47.0	100.0	✗
#21	chini-train-08	base_3b single-shot	73	56.0	47.0	100.0	✗
#22	chini-train-08	base_3b single-shot	73	56.0	47.0	100.0	✗
#23	chini-train-08	base single-shot	73	56.0	47.0	100.0	✗
#24	chini-train-08	fmt_a_v2 single-shot	71	52.0	42.0	100.0	✗
#25	chini-train-08	fmt_a_3b single-shot	71	52.0	42.0	100.0	✗
#26	chini-train-08	fmt_a_3b single-shot	71	52.0	42.0	100.0	✗
#27	chini-train-08	base_7b single-shot	57	62.0	41.0	70.0	✗
#28	chini-train-08	base_7b single-shot	57	62.0	41.0	70.0	✗
#29	chini-train-08	base_7b single-shot	57	62.0	41.0	70.0	✗
#30	chini-train-08	base single-shot	57	62.0	41.0	70.0	✗

Per-scenario breakdown of the top run

Scenario	Health	Drop rate	Delivered	Pass
baseline	75.0	2.6%	57	✗
adversarial-1	49.0	100.0%	300	✗
outage-2	66.0	5.7%	33	✗
spike-3	63.0	8.8%	3217	✗
cascade-4	17.0	61.8%	17	✗
latency-5	71.0	4.3%	66	✗

How is this scored? →