Visual Time Perception & Symbolic Reasoning - Single & Continuous Conversion

Architecture, System Design, Conceptual Framework, and Core Development by Franz Ayestaran / Enhanced Pair Programming with OpenAI GPT

Theme

Live Processing Histogram

Angle bins (0-360) and mask pixel-count diagnostics (not time units) update every symbolic conversion.

Randomize era theme every second Increment clock rotation every second Random flip clock every second 3D rotate with back view Extra distractor line Artifact-safe mode

Symbolic Output Probabilistic Logic Reasoning Symbolic Output Log

Grounding Preset Studio

Compose reusable symbolic grounding strategies by dragging presets into the script generator. Apply rules to current symbolic output, copy as JSON, or clear to start over. Scripts persist across page refresh.

Preset Library

Script Generator

Drop presets here to build a grounding script.

Generated Grounding Script

Generated Dataset

Accumulate labeled samples from symbolic frames. Build dataset, filter by search or instruction type, then copy or save as JSONL. Rows persist until cleared.

No dataset generated yet.

Reproducible Experiments – Run Sweep

Execute a full robustness sweep with 10 degradation conditions (rotate, blur, noise, occlusion, mirror, etc.). Generates synthetic clock images, tests symbolic detection, and runs statistical significance tests. Condition sample thumbnails appear during the run and persist in browser cache.

Samples/condition Stats iterations

No experiment run yet.

Idle

Queued Sampling Stats Done

Stats Summary Preview

LLM Benchmark Comparison

Compare multiple LLM models on the same set of clock-reading questions (requires local Ollama). Generates questions from symbolic experiment metrics, collects responses, scores correctness and key-factor extraction, and runs two-way ANOVA + Tukey HSD statistical tests. Parse success summary shows JSON parse rates per model.

Models Questions/condition Timeout (s) 0 selected

No LLM benchmark run yet.

Idle

Queued Questions Responses Evaluation Stats Done

Available LLM Artifacts

Filter Sort

LLM Benchmark Stats Summary