LAB / WMSI-EXP-072
Vocabulary Suppression in Specialized Models.
Does domain-shaped framing persist when obvious domain vocabulary is forbidden?
01 / QUESTION
Representation Search and Specialized Intellectual Geometry
Does domain-shaped framing persist when obvious domain vocabulary is forbidden?
This is a protocol scaffold, not a completed experiment or a frozen execution plan. No measured effect or research conclusion is claimed here.
Hypothesis to test
The manipulation described below produces a reproducible, condition-specific change in the measured system behavior or internal state that is not explained by the ordinary controls.
Ordinary competing explanation
The observed pattern is explained by ordinary sampling variation, prompt/representation artifacts, surface lexical cues, evaluator bias, or the known product/runtime mechanics captured by the controls.
02 / PLANNED DESIGN
How we intend to test it
Repeat specialized-model tests while banning signature terms such as energy, selection, equilibrium, utility, etc.
Controls
Paraphrase controls; generalist model with same word bans.
What we would measure
Structural similarity to unrestricted answers, lexical substitution, causal graph overlap.
Access and instruments
Local models or internal-state access. This describes access requirements, not verified readiness on our Spark.
Fine-tuned models; constrained decoding/prompting.
Before a full run
Feasibility=LOCAL; priority=P0. Before confirmatory execution, run a small pilot to estimate variance/cost, freeze the scoring rules, and assign a run-count target.
Model revisions, final prompts, sample sizes, cost limits, and scoring rules must be settled through the pilot and protocol-freeze process.
03 / INTERPRETATION
What could fool us
Different data quality, vocabulary leakage, compute mismatch, representation chosen after seeing answer.
What would weaken the idea
If framing vanishes with terminology, the effect is mostly metaphorical surface transfer.
Observation, inference, speculation, unknowns, and apparatus errors will be labeled separately. One striking transcript cannot establish a general effect.
04 / RECORD
Evidence we plan to preserve
Base checkpoint, training corpus hash, recoding candidates, dependency graphs, solution traces.
Planned visual evidence
Graph-overlap plots; banned-term replacement map.
No research runs or raw datasets are published for this experiment. The September 26, 2026 registry and protocol compendium (version 1.0) are the sources for this planning record.
Back to all 90 experiments