TwoQuarks Research.

Empirical evaluation. Cross-model evaluation of isomeric polarization, graph-momentum diagnostics, and inference-time instability across production LLM APIs, using provider-agnostic black-box methods.

Papers ↓ CRAD → GitHub
p = 0.0013
C2 Refusal Erosion · ρ = +0.713 · Claude Haiku
L₃ = 0.000
Control-negative · both models
17
Context depths · seed-controlled
2
Models · Claude · GPT
Method

Cross-model PfV evaluation

Polarization-from-Views (PfV) estimates a black-box proxy for structural divergence across multiple realizations of a model response. In this series, it was evaluated on Claude Haiku and GPT-4o-mini against a 5,000-permutation null; deterministic controls produced L₃ = 0.000 in both.

Primary result

C2 Refusal Erosion

The strongest observed signal in this series: refusal erosion in Claude Haiku correlates with ΔL₃ at ρ = +0.713 (p = 0.0013), seed-controlled across 17 context depths. C3 Anchor Displacement is a promising cross-model candidate (ρ = +0.799 GPT-mini, +0.647 GPT-full) but remains marginal in the pooled cross-model test (p = 0.054) and is reported as a direction, not a claim. Probe cases C1–C5 map sycophancy, refusal erosion, anchor displacement, rule override, and reasoning drift onto separate signal channels.

Open question

Is inference-time stability regulation sufficient to prevent unsafe behavior under regime shift?

TwoQuarks treats model instability as drift and regime transition rather than only as a final unsafe answer. The research program asks whether monitoring and lightweight intervention at inference time — leaving parameters, policies, and training objectives untouched — can identify or mitigate candidate precursors before output-level failure. Current results are treated as bounded empirical observations, not universal predictors.

White-box companion

CRAD — the mechanistic arm →

The evaluation above is provider-agnostic and black-box. Its white-box counterpart, Constraint-Retained Action Divergence, reads the residual stream of a frozen model directly and documents a pre-leak candidate where the action diverges while the safety constraint stays retained — a deterministic case study with explicit falsification criteria.

Architecture research · active prototype

Dopamine — the architecture arm →

Dopamine is the constructive branch of the program: an experimental small language model that separates multi-scale representation, latent integration, memory pathways, and output transmission into inspectable components. Current evidence is developmental rather than comparative; the page reports the recorded run and the missing controls explicitly.

Papers

Open to read, cite, and reproduce. All work is authored at TwoQuarks Research.

Paper · 2026

A Modular Framework for Adaptive Stability Control in Sequence Models Under Regime Uncertainty

J. Ledesma · 2026 · TwoQuarks Research, preprint

The inference-time control architecture and the six-flavor formulation: monitoring and lightweight intervention that leave parameters, policies, and training objectives untouched.

PDF ↗

Paper · 2026

Constraint-Retained Action Divergence: Neutral-Anchored Detection of Pre-Leak State–Action Dissociation in a Frozen Language Model

L. J. Ledesma Pérez · 2026 · Zenodo, TwoQuarks Research

White-box mechanistic detection of a pre-leak candidate where the action diverges while the safety constraint stays retained in the residual stream. A deterministic case study with explicit falsification criteria.

DOI 10.5281/zenodo.21315238 ↗  ·  CRAD page →

Paper · 2026

Molecule Model: Isomeric Polarization and Adaptive Stability Control — Detecting and Mitigating Structural Reorganization in Large Language Models

J. Ledesma · 2026 · TwoQuarks Research, preprint

The first operational instrument: a six-flavor black-box instability monitor that turns the polarization diagnostics into a running measurement.

PDF ↗

Paper · 2026

Isomeric Polarization: Internal Structural Divergence as Emergent Property of Computational Systems

J. Ledesma · 2026 · Zenodo, TwoQuarks Research

The PfV metrics (L1, L2, ΔL₃) and the polarization formulation behind the diagnostics — internal structural divergence framed as an emergent property.

DOI 10.5281/zenodo.19675750 ↗  ·  PDF ↗

Executive summary — one-document overview ↗

# reproducible — install the pipeline and probe any LLM from your own code pip install twoquarks # framework indexed on Zenodo · DOI 10.5281/zenodo.19675750