Figure Inventory¶

Source of truth: figure_inventory.json. This document is a human-readable view — edit the JSON, not this table.

Three distinct categories of figures are produced by three separate generators. Each generator writes to its own directory; they do not cross-populate each other's directories by default.

Category roles¶

Category	Generator	Output directory	Format	Purpose
`didactic`	`scripts/generate_docs_images.py`	`docs/images/`	PNG	API docs, README, docs site
`evidence`	`scripts/generate_research_figures.py`	`docs/images/research/`	PNG + PDF	Research docs, benchmark evidence
`paper`	`scripts/generate_paper_figures.py`	`paper/figures/`	PDF	`paper/main.tex` only

To place evidence figures inside paper/figures/, invoke the generator explicitly with --pdf-dir paper/figures. This is not done automatically.

Regeneration commands¶

make docs-images       # didactic figures only
make research-figures  # evidence figures only
make paper-figures     # paper figures only
make figures           # all three

Didactic figures (`docs/images/`)¶

Stem	Plot kind	Description	Referenced in
`circumplex`	scatter	Russell (1980) circumplex with sample memory positions	README.md
`decay_curves`	line	ACT-R power-law decay at three arousal levels	README.md
`yerkes_dodson`	line	Yerkes-Dodson arousal-performance curve	README.md
`retrieval_radar`	radar	Six-signal retrieval score breakdown	README.md
`mood_evolution`	line_multipanel	PAD mood-field trajectory over 30 minutes	README.md
`adaptive_weights_heatmap`	heatmap	Adaptive weight distribution across valence/arousal space	README.md
`resonance_network`	network_graph	Resonance link graph with five link types	README.md
`appraisal_radar`	radar	Scherer CPM five-dimensional appraisal vector	README.md

All didactic figures use synthetic/hardcoded inputs — no benchmark JSON.

Evidence figures (`docs/images/research/`)¶

Stem	Plot kind	Source benchmark(s)	Claim	Referenced in
`research_realistic_v2_overview`	bar_with_ci	`benchmarks/realistic/results.v2.sbert.json`	`replayable_multi_session_help`	README.md, `09_current_evidence.md`
`research_realistic_v2_challenges`	grouped_bar	`benchmarks/realistic/results.v2.sbert.json`	`replayable_multi_session_help`	README.md, `09_current_evidence.md`
`research_ablation_s3`	bar_forest	`benchmarks/ablation/results.v2.sbert.json`	`theory_faithful_operationalization`	README.md, `09_current_evidence.md`
`research_multilingual`	grouped_bar_with_ci	`results.v2_{it,es}.{sbert,me5}.json` (N=80) + `results.v2_fr.me5.json` (N=120 Hm1)	`retrieval_affect_aware`, `cross_domain_affect_replication`	README.md, `09_current_evidence.md`
`research_locomo_negative`	grouped_bar_with_ci	`benchmarks/locomo/results.json`	`locomo_external_qa_negative`	README.md, `09_current_evidence.md`
`research_third_party_x_x2`	grouped_bar	`madialbench/results.json` + `esmemeval/results.json`	`cross_domain_affect_replication`	README.md, `09_current_evidence.md`
`research_query_affect_gate`	grouped_bar	`benchmarks/gate/results.json`	`query_affect_gate`	README.md, `09_current_evidence.md`
`research_learned_profile`	grouped_bar	`benchmarks/learned_profile/results.json`	`cross_domain_affect_replication`	README.md, `09_current_evidence.md`

Claim IDs map to docs/research/claim_validation_matrix.json.

Paper figures (`paper/figures/`)¶

Stem	Plot kind	Description	Referenced in
`fig1_circumplex`	scatter	Russell circumplex — 12 seeded memories	`paper/main.tex`
`fig2_decay_curves`	line	ACT-R decay modulated by arousal	`paper/main.tex`
`fig3_mood_evolution`	line_multipanel	PAD trajectory over simulated conversation	`paper/main.tex`
`fig4_resonance_network`	network_graph	Resonance link graph snapshot	`paper/main.tex`

Paper figures are schematic/synthetic (seeded random data); they do not reference benchmark JSON.

Planned evidence figures (Fase 2)¶

The following benchmark JSON artefacts exist but are not yet visualised:

Planned stem	Source JSON	Claim
`research_sota_realistic_v2`	`benchmarks/comparative/results.sota.v2.sbert.json`	`realistic_replay_vs_sota`
`research_dailydialog_negative`	`benchmarks/dailydialog/results.json`	`cross_domain_affect_replication`
`research_hi3_resonance`	`benchmarks/ablation/results.hi3.json`	`resonance_amplification_e5`
`research_hd2_multilingual_power`	`benchmarks/appraisal_confound/results.hd2_{it,es}.me5{,.v120}.json`	`retrieval_affect_aware`
`research_locomo_pareto`	`benchmarks/locomo/pareto_results.json`	`locomo_external_qa_negative`