Best LLMs for Community Policy Evaluation
Determines whether supplied community rules permit a proposed action under caller-defined channel and decision policies, returning a conservative decision and evidence. Illustrative uses include deciding whether a product update fits a customer group, an open-source reply follows
Run this task on Fronset — request an invitation
Models
Frontier on this task: MiniMax M3 at 8.47 / 10. Quality bar at 90%: 7.63.
point-estimate floor (CI low) · upper CI (less certain) · Bars sorted by blended cost; best-value model first.
| Model | Quality score | CI low | Cost / 1k runs | vs best value |
|---|---|---|---|---|
| MiniMax M3 | 8.47 / 10 | 8.25 | $0.40 | best value |
Cost breakdown
| Model | Quality | Confidence | Cost / 1k runs | Overpay | Mode |
|---|---|---|---|---|---|
| MiniMax M3 ★ best OpenRouter | 8.47 / 10 CI [8.25, 8.70] | HIGH | $0.40 | best value | batch |
Overpay shows how much more you pay than the best-value model that clears the quality bar (marked ★) — the best-value good-enough option. "16x" means you overpay 16× — 16× that reference for no quality benefit above the bar. Typical call shape for this task: 1429 input tokens → 560 output tokens, EMA-tracked from production traffic. Cost is the observed, all-in $ per 1,000 task runs: each model's own measured usage on this task — output verbosity, thinking/reasoning tokens, cache reads and writes, and the spend on its billed failures — priced at current list rates and adjusted by the billing overhead we actually reconcile against provider invoices. Models that answer tersely cost what they actually cost; models that think at length pay for it. Not comparable to providers' advertised $/1M list rates — this is what running the task costs, not a per-token price.
Evaluation rubric
Judge rule interpretation, evidence traceability, distinction among automation and other restrictions, decision calibration under incomplete rules, and correct identification of conditions. Do not judge legal compliance beyond the supplied policy.
Output schema
Every answer on this task is checked against this JSON Schema, whichever model wrote it. An answer that doesn't fit counts as a model failure, and the call is retried on another model.
{
"properties": {
"is_posting_allowed": {
"description": "Whether this subreddit is suitable for posting analytical content",
"title": "Is Posting Allowed",
"type": "boolean"
},
"posting_rules_summary": {
"description": "Concise summary of the subreddit's posting rules relevant to financial analysis content (2-4 sentences)",
"title": "Posting Rules Summary",
"type": "string"
},
"reasoning": {
"description": "Why this subreddit is or is not suitable for posting",
"title": "Reasoning",
"type": "string"
}
},
"required": [
"is_posting_allowed",
"posting_rules_summary",
"reasoning"
],
"title": "SubredditVettingResult",
"type": "object"
}Prompt templates
The published template for this task.
default ·
LLMB_COMMUNITY_POLICY_EVALUATION_SYSTEM +
LLMB_COMMUNITY_POLICY_EVALUATION_USER
(0 calls in window)
System prompt
Apply only the supplied community, community_rules, proposed_action, channel_profile, and policy_profile. Use the profiles as the sole source of rule concepts, eligibility conditions, evidence standards, and decision behavior. Return conditional or unknown when the supplied rules are incomplete and reference the decisive rule text. Treat empty optional values as absent and return only the requested result. Your response must conform exactly to this output schema: {schema_json_string}.
User prompt
Inputs — community: {community}; community_rules: {community_rules}; proposed_action: {proposed_action}; channel_profile: {channel_profile}; policy_profile: {policy_profile}. Use only these inputs to complete the task defined by the system prompt.