run record~/research/cliste/runsAll Cliste runs

Type a section name and press Enter. Use “help” for the list.

Figures: softened for dark viewingDisplay only; image files are unchanged.
← All Cliste runs

Cliste experiment

Recognize sound sequences through retained temporal content

Execution: completedScientific verdict: partial

The question

Test whether four disjoint acoustic blocks distinguish episode recognition from similarity.

Implementation and comparison details

Frozen encoder, vector-derived temporal paths, alternative alignments and explicit order evidence; prior 108 encounters preserved in a cloned life.

What happened

All bounded usability and retention floors pass: 213/216 moderate correct, 1/100 environmental false recognition, no shuffled/repeated sequence acceptance, and all 108 old exact memories preserved. Overall partial: challenge gain 8.3 percentage points misses the registered 15-point target; strong noise remains weak.

Results and visual artifacts

Open a figure for its input, comparison, and result. Click an image to enlarge it.

Sequence recognition and false claims
Sequence recognition and false claims
Matched fresh recordings and 300 retained encounters. The unordered control uses the same four sound blocks; alternative alignments never add independent votes. Single-window acceptance is shown as an insufficient sequence-identity proxy.
bunny-03: white20
bunny-03: white20
Actual native vector decoding, including refused or wrong candidates. Observer input is separate from retained memory. Attribution: Blender Foundation, Big Buck Bunny / Sintel / Elephants Dream (peach.blender.org, durian.blender.org, orange.blender.org).
bunny-03: hum20
bunny-03: hum20
Actual native vector decoding, including refused or wrong candidates. Observer input is separate from retained memory. Attribution: Blender Foundation, Big Buck Bunny / Sintel / Elephants Dream (peach.blender.org, durian.blender.org, orange.blender.org).
bunny-02: white20
bunny-02: white20
Actual native vector decoding, including refused or wrong candidates. Observer input is separate from retained memory. Attribution: Blender Foundation, Big Buck Bunny / Sintel / Elephants Dream (peach.blender.org, durian.blender.org, orange.blender.org).
sintel-01: white20
sintel-01: white20
Actual native vector decoding, including refused or wrong candidates. Observer input is separate from retained memory. Attribution: Blender Foundation, Big Buck Bunny / Sintel / Elephants Dream (peach.blender.org, durian.blender.org, orange.blender.org).
One remaining false episode recognition
One remaining false episode recognition
The environmental sound was never enrolled, yet the system selected a retained Dune sequence. Observer spectra illustrate possible resemblance; neither category recognition nor waveform reconstruction from VSA is claimed.

Limits and next question

Uniform supplied two-second boundaries; three fresh open-film sources and 50 environmental files. No semantic classes, autonomous segmentation or hardware qualification. Skala is not built.

Next question. Can weak evidence within a sequence be handled without discarding valid order or admitting more false episodes? Preserve this qualified bounded behavior as the comparison.

Inspect the evidence

Download the methods and measurements for this experiment. Large training and saved-state files are not included.