Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/elder-plinius/OBLITERATUS
/ functions
Functions
1,841 in github.com/elder-plinius/OBLITERATUS
⨍
Functions
1,841
◇
Types & classes
359
Method
test_direction_specificity_test
(self)
tests/test_module_imports.py:50
Method
test_direction_stats
(self)
tests/test_telemetry.py:157
Method
test_directions_are_unit_vectors
Per-expert directions should be unit normalized.
tests/test_abliterate.py:2163
Method
test_directions_are_unit_vectors
Extracted directions should be unit length.
tests/test_analysis.py:40
Method
test_disable_programmatically
(self)
tests/test_telemetry.py:79
Method
test_disable_via_env_false
(self)
tests/test_telemetry.py:65
Method
test_disable_via_env_zero
(self)
tests/test_telemetry.py:60
Method
test_disabled_by_default
(self)
tests/test_telemetry.py:46
Method
test_distributed_detects_not_sft
Distributed refusal should not be predicted as SFT.
tests/test_novel_analysis.py:249
Method
test_does_not_send_when_disabled
(self)
tests/test_telemetry.py:328
Method
test_does_nothing_when_disabled
(self)
tests/test_telemetry.py:421
Method
test_does_nothing_when_disabled
(self)
tests/test_telemetry.py:521
Method
test_double_projection_is_idempotent
Projecting twice should give the same result as projecting once.
tests/test_abliteration_math.py:86
Method
test_dpo_zero_regularization
(self)
tests/test_informed_pipeline.py:177
Method
test_effective_rank
Multi-direction subspace should yield effective rank > 1.
tests/test_telemetry.py:235
Method
test_effective_rank_change
Should compute before/after rank comparison.
tests/test_advanced_metrics.py:180
Method
test_effective_rank_nan_matrix
effective_rank should handle matrix with NaN.
tests/test_edge_cases.py:111
Method
test_ega_distill_integration
EGA should be called during distill when per_expert_directions is enabled.
tests/test_abliterate.py:2553
Method
test_ega_only_on_primary_direction
EGA should only apply for dir_idx==0, not higher SVD directions.
tests/test_abliterate.py:2543
Method
test_ega_pipeline_flags
Pipeline with surgical method should enable per_expert_directions.
tests/test_abliterate.py:2538
Method
test_ega_with_routing_data_end_to_end
End-to-end: EGA computes directions and granular projection modifies weights.
tests/test_abliterate.py:2578
Method
test_emit_records_durations
Verify _emit stores durations in _stage_durations dict.
tests/test_telemetry.py:579
Method
test_empty
(self)
tests/test_metrics.py:50
Method
test_empty_activations
(self)
tests/test_novel_analysis.py:202
Method
test_empty_batch
(self)
tests/test_novel_analysis.py:472
Method
test_empty_categories
(self)
tests/test_breakthrough_modules.py:378
Method
test_empty_directions
Should handle empty input gracefully.
tests/test_logit_lens.py:140
Method
test_empty_directions
(self)
tests/test_novel_analysis.py:281
Method
test_empty_directions
(self)
tests/test_telemetry.py:246
Method
test_empty_input
(self)
tests/test_abliterate.py:1666
Method
test_empty_input
Should handle empty input gracefully.
tests/test_analysis.py:192
Method
test_empty_input
Should handle empty input gracefully.
tests/test_analysis.py:340
Method
test_empty_input
(self)
tests/test_breakthrough_modules.py:173
Method
test_empty_input
(self)
tests/test_breakthrough_modules.py:280
Method
test_empty_input
(self)
tests/test_breakthrough_modules.py:517
Method
test_empty_input
(self)
tests/test_new_analysis_modules.py:624
Method
test_empty_inputs
(self)
tests/test_novel_analysis.py:640
Method
test_empty_list
(self)
tests/test_heretic_eval.py:89
Method
test_empty_list
(self)
tests/test_refusal_detection.py:98
Method
test_empty_list
(self)
tests/test_analysis_utils.py:13
Method
test_empty_model_name
Empty string should fall through to Dense Standard.
tests/test_architecture_profiles.py:551
Method
test_empty_overrides
(self)
tests/test_architecture_profiles.py:399
Method
test_empty_pipeline
(self)
tests/test_defense_robustness.py:74
Method
test_empty_responses
(self)
tests/test_refusal_detection.py:133
Method
test_empty_responses
(self)
tests/test_advanced_metrics.py:73
Method
test_empty_responses_not_refusals
Per Zou et al. logic, empty responses have no substring match = not refusals.
tests/test_heretic_eval.py:84
Method
test_empty_string
(self)
tests/test_refusal_detection.py:23
Method
test_empty_string_not_refusal
(self)
tests/test_advanced_metrics.py:76
Method
test_enable_programmatically
(self)
tests/test_telemetry.py:75
Method
test_enable_via_env_explicit
(self)
tests/test_telemetry.py:70
Method
test_enabled_by_default_on_hf_spaces
(self)
tests/test_telemetry.py:51
Method
test_energy_removed
Top rows should capture most of the refusal energy.
tests/test_novel_analysis.py:523
Method
test_entangled_layers_skipped
(self)
tests/test_informed_pipeline.py:250
Method
test_entanglement_gate
(self, pipeline)
tests/test_informed_pipeline.py:117
Method
test_entanglement_map
(self)
tests/test_defense_robustness.py:135
Method
test_entropy_nonnegative
Logit effect entropy should be non-negative.
tests/test_logit_lens.py:154
Method
test_enumerate
(self, handle)
tests/test_strategies.py:77
Method
test_enumerate
(self, handle)
tests/test_strategies.py:112
Method
test_enumerate
(self, handle)
tests/test_strategies.py:141
Method
test_enumerate
(self, handle)
tests/test_strategies.py:162
Method
test_environment_info
(self)
tests/test_telemetry.py:145
Function
test_estimate_total_params_uses_nested_text_config
()
tests/test_model_profile.py:11
Method
test_estimated_passes
(self)
tests/test_breakthrough_modules.py:287
Method
test_excise_advanced_norm_preserving
Advanced method with norm preservation should maintain weight norms.
tests/test_abliterate.py:1843
Method
test_excise_basic
Basic method should modify weights.
tests/test_abliterate.py:1809
Method
test_excise_details
(self)
tests/test_telemetry.py:162
Method
test_expected_benchmarks
(self)
tests/test_heretic_eval.py:305
Method
test_expert_input_projections_projected
Expert input projections (up_proj, gate_proj) should also be modified.
tests/test_abliterate.py:527
Method
test_expert_transplant
Expert transplant should overwrite safety expert weights with capability average.
tests/test_abliterate.py:1552
Method
test_explicit_layer_mapping
(self)
tests/test_breakthrough_modules.py:477
Method
test_explicit_overrides_method
(self)
tests/test_abliterate.py:203
Method
test_explicit_overrides_new_params
(self)
tests/test_abliterate_extended.py:127
Method
test_extract_all_layers
Should extract directions for all provided layers.
tests/test_analysis.py:80
Method
test_extract_all_layers
(self)
tests/test_new_analysis_modules.py:463
Method
test_extract_number
(self)
tests/test_benchmarks.py:176
Method
test_falcon3_is_not_reasoning
falcon3 must NOT match 'o3' reasoning pattern.
tests/test_architecture_profiles.py:167
Method
test_fallback_to_next_key
(self)
tests/test_heretic_eval.py:357
Method
test_fallback_to_shared_on_fused
Experts without per-expert direction should use shared direction.
tests/test_abliterate.py:2445
Method
test_fallback_when_no_template
Should fall back to raw prompts when template is not configured.
tests/test_abliterate_extended.py:244
Method
test_falls_back_to_shared_direction
Experts without per-expert direction should use shared direction.
tests/test_abliterate.py:2292
Method
test_false_positive_sorry
I'm sorry to hear that' followed by helpful content.
tests/test_edge_cases.py:300
Method
test_feature_clustering
(self)
tests/test_new_analysis_modules.py:391
Method
test_feature_extraction
(self)
tests/test_novel_analysis.py:269
Method
test_fewshot_settings_match_arditi
Few-shot settings should match Arditi et al. / Open LLM Leaderboard.
tests/test_heretic_eval.py:309
Method
test_filename_format
(self, tmp_path)
tests/test_community.py:155
Method
test_filters_unknown_config_keys
(self)
tests/test_telemetry.py:124
Method
test_finds_gate
Should find a router named 'gate'.
tests/test_abliterate.py:1958
Method
test_finds_router
Should find a router named 'router'.
tests/test_abliterate.py:1972
Method
test_first_token_kl_different
First-token KL of different distributions should be positive.
tests/test_advanced_metrics.py:129
Method
test_first_token_kl_identical
First-token KL of identical distributions should be 0.
tests/test_advanced_metrics.py:123
Method
test_five_percent_threshold_includes_more
Layers between 5% and 10% of max should now be included.
tests/test_abliterate.py:434
Method
test_five_standard_benchmarks
(self)
tests/test_heretic_eval.py:296
Method
test_flat_manifold_detection
When activations are purely linear, curvature should be near zero.
tests/test_breakthrough_modules.py:189
Method
test_format_analysis
(self)
tests/test_novel_analysis.py:619
Method
test_format_cross_category
(self)
tests/test_causal_and_transfer.py:497
Method
test_format_cross_model
(self)
tests/test_causal_and_transfer.py:488
Method
test_format_decomposition
(self)
tests/test_causal_and_transfer.py:254
Method
test_format_default
(self, insights)
tests/test_informed_pipeline.py:302
Method
test_format_entanglement
(self)
tests/test_defense_robustness.py:163
Method
test_format_imprint
(self)
tests/test_novel_analysis.py:307
← previous
next →
1,201–1,300 of 1,841, ranked by callers