10 Services from aialign.halowerk.com

Computes pairwise action-agreement statistics (observed, expected, and Cohen's kappa) across multi-agent rounds to flag suspiciously high coordination above a configurable threshold.

Detects sycophantic drift in AI responses by comparing cosine similarity between a leading user position and paired independent vs. conditioned responses, reporting the positive similarity shift.

Calculates positive-outcome rates per group, compares each against a reference group, and reports rate differences and selection-rate ratios to surface potential disparate impact.

Computes Brier loss, expected calibration error (ECE), and maximum calibration error (MCE) from a set of confidence scores and binary correctness labels, partitioned into equal-width bins.

Scans a prompt against defensive regex categories and returns category labels, a bounded risk score, and a block/review recommendation.

Computes cosine similarity and normalized L1 difference between initial and current goal weight vectors, flagging drift when cosine similarity falls below a supplied threshold.

Tokenizes claims and evidence, removes English stop words, and computes the fraction of unique claim tokens found in evidence, flagging claims below a coverage threshold.

Redacts private keys, credentials, emails, payment card numbers, IPv4 addresses, and phone-like digit strings from text, returning sanitized content with per-category match counts.

Divides observed resource usage (CPU, memory-time, network, tool calls, wall time) by caller-supplied budgets and ranks agents by their worst overrun ratio

Aggregates caller-labeled test cases into weighted pass rates and weighted scores with deterministic per-category summaries for AI capability evaluation.