Added numbers to BertViz visualization, easier to map attention head numbers

added color swatches for easy mapping from attention categories to BertViz

Fixed model-switching bug with activation store clearing

Attention refactor, better categorization and explanation

ablation experiment fixes, better UI and full sequence comparison

feat(output): Add token scrubber with per-position top-5 next-token probabilities

chore: remove unused code, imports, and deprecated functions

conductor setup

conductor(checkpoint): Checkpoint end of Phase 3: Visualization & Feedback Loop

conductor(checkpoint): Checkpoint end of Phase 2: Frontend Control Panel

feat(ablation): Unify ablation execution into execute_forward_pass

Fix multi-layer ablation to ablate all heads simultaneously

Switch attention viz to head_view and add categorization (Agent D)

heatmap implementation

Ablation updated for full sequence, needs refactor for front-end and workflow

Fixed errors for beam search refactor

Fix meta tensor error in safe_to_serializable

Update layer highlighting to 100% threshold for actual output token only

Clean up: fix significant layer threshold to 50%, remove unused logit lens parameter, remove certainty calculations

Fix ablation: add attention_modules to session store so ablation can find attention layers

Add comprehensive debugging to ablation data flow

Fix ablation: capture ablated attention module output to populate layer data

Reorganize Layer-by-Layer Predictions: move diagram to main container, reorder sections, update arrows, increase threshold to 75%

Fix token predictions not being collected - changed condition to check for block_outputs and norm_parameters instead of logit_lens_parameter

Phase 1: Backend infrastructure for token merging, layer-wise tracking, and head ablation

Fix experiments section visibility and improve ablation logic - Fix experiments section callback to show properly after analysis - Change mean ablation to use mean across ALL layers and tokens (not just selected layer) - Update UI description to clarify meaningless mean activation - Remove prevent_initial_call to ensure proper initialization

Add Experiments panel with mean ablation feature - Remove Divergent Layers summary from comparison panel (keep badges) - Add collapsible Experiments section after analysis - Implement Ablation Study with layer selection buttons - Add mean ablation function to replace layer activations with mean - Run ablation on both prompts and update results - Gray out ablated layer in accordion with visual indicator - Add CSS styling for experiments panel and ablated layers

Fix meta tensor serialization error in safe_to_serializable

Add attention head categorization with colored badges to each layer accordion

Fix attention view display: always show section, remove Analysis Results, add debugging

Add simplified attention view with top-3 attended tokens and BertViz toggle button

feat: Add top-5 predictions with deltas, certainty meter, and bar charts in accordion panels

Fix accordion data flow and remove all legacy Cytoscape code

Replace Cytoscape with accordion-based layer view

Refactor Check Token: Replace 4th edge with probability line graph

Implement edge hover tooltips with floating divs for better UX

Replace edge hover info with tooltip bubbles on visualization

Fix BertViz category visualization tensor stacking error and add head index labels

Feature update: Reorganize attention head categorization with BertViz visualizations for each category

Feature update: Check token now tries both with/without leading space and uses higher probability variant

Feature 4: Replace BertViz head_view with model_view for hierarchical attention visualization

Feature updates: collapsible sidebar, toggle compare button, check token submit button, fix second graph visibility, improve check token debug output

Add check token feature with 4th edge and minimum opacity

Using final layer output instead of mlp for simpler computation

simplified vocab space projection

Added model config for automatic model parameter detection

Uses pyvene instead of just pytorch hooks, setting up for activation patching

Full visualization reorganization, handles model selection, parameter selection, visualizes graphs w/ bertviz onclick