MCPcopy Create free account

hub / github.com/adamkarvonen/activation_oracles / functions

Functions716 in github.com/adamkarvonen/activation_oracles

⨍ Functions 716◇ Types & classes 122
↓ 1 callersMethodresponse_with_logprobs
(self)
nl_probes/autointerp_detection_eval/caller.py:351
↓ 1 callersFunctionrun_async_evaluation
()
nl_probes/autointerp_detection_eval/local_eval.py:293
↓ 1 callersFunctionrun_eval_for_datasets
Returns: results[dataset_id][method_key] -> metrics dict
experiments/classification_eval.py:215
↓ 1 callersFunctionrun_evaluation_for_explanations
Run precision/recall evaluation for a set of explanations using their hard negatives.
nl_probes/autointerp_detection_eval/eval_detection_v2.py:639
↓ 1 callersFunctionrun_explanation_generation_hf
Calls your existing evaluation loop that produces per-feature explanations.
nl_probes/autointerp_detection_eval/lora_hf_eval.py:129
↓ 1 callersFunctionrun_gemma_steering
Run gemma steering for a single SAE.
nl_probes/autointerp_detection_eval/eval_detection_v2.py:871
↓ 1 callersFunctionrun_gemma_steering_best_of_n
Run gemma steering for a single SAE with best-of-n.
nl_probes/autointerp_detection_eval/eval_detection_v2.py:910
↓ 1 callersFunctionrun_single_prompt
Run a single game and return the model's response.
nl_probes/autointerp_detection_eval/caller.py:1062
↓ 1 callersFunctionrun_voting
(queries, args)
datasets/latentqa_datasets/curate_gpt_data.py:175
↓ 1 callersFunctionscore_detection
( prompts: list[eval_detection_v2.SAETrainTestWithExplanation], detection_model: str = "gpt-5-mini-202
nl_probes/autointerp_detection_eval/lora_hf_eval.py:249
↓ 1 callersFunctionscore_eval_responses
( eval_responses: list[FeatureResult], eval_dataset: list[TrainingDataPoint], valid_answers: list[
nl_probes/utils/eval.py:166
↓ 1 callersFunctionsentence_to_prompt_with_vector
Convert a SentenceInfoV2 object to a prompt.
nl_probes/autointerp_detection_eval/eval_detection_v2.py:135
↓ 1 callersFunctionsort_key
(series_name: str)
experiments/final_paper_plots/plot_model_progression_line_chart_shapes.py:812
↓ 1 callersMethodsorted_logprobs
(self)
nl_probes/autointerp_detection_eval/caller.py:225
↓ 1 callersFunctionssc_analyse_quirk
( records: list[SscRecord], response_type: ResponseType = "token_responses", best_of_n: int = 5 )
experiments/final_paper_plots/plot_model_progression_line_chart_shapes.py:604
↓ 1 callersFunctionssc_extract_message_llm
(text: str)
experiments/final_paper_plots/plot_model_progression_line_chart_shapes.py:581
↓ 1 callersFunctionssc_get_best_of_n_scores
Calculate best-of-N scores for each record, filtered by the best investigator prompt.
experiments/final_paper_plots/plot_model_progression_line_chart_shapes.py:631
↓ 1 callersFunctionssc_judge_quirk
(ground_truth: str, response: str, response_type: ResponseType)
experiments/final_paper_plots/plot_model_progression_line_chart_shapes.py:547
↓ 1 callersFunctionssc_load_json_schema
(json_path: str)
experiments/final_paper_plots/plot_model_progression_line_chart_shapes.py:669
↓ 1 callersFunctionssc_load_results
Load all JSON files from the directory.
experiments/final_paper_plots/plot_model_progression_line_chart_shapes.py:675
↓ 1 callersFunctionstandardize_train_test
(X_tr: torch.Tensor, X_te: torch.Tensor)
experiments/linear_probe.py:156
↓ 1 callersFunctionsummarize_detection_scores
(results, verbose: bool = False)
nl_probes/autointerp_detection_eval/lora_hf_eval.py:276
↓ 1 callersFunctiontaboo_calculate_accuracy
(record: dict, investigator_lora: str | None, sequence: bool)
experiments/final_paper_plots/plot_qwen3-8b_eval_results.py:120
↓ 1 callersFunctiontaboo_calculate_accuracy
( record: dict, investigator_lora: str | None, sequence: bool, model_name: str | None = None )
experiments/final_paper_plots/plot_secret_keeping_results.py:107
↓ 1 callersFunctiontaboo_calculate_accuracy
( record: dict, investigator_lora: str | None, sequence: bool, model_name: str | None = None )
experiments/final_paper_plots/plot_model_progression_line_chart_shapes.py:323
↓ 1 callersFunctiontaboo_calculate_accuracy
(record: dict, investigator_lora: str, sequence: bool)
experiments/plotting/plot_secret_keeping_results.py:99
↓ 1 callersFunctiontest_response
Generate multiple test responses for a given message.
experiments/simple_llama_model_demo.py:84
↓ 1 callersFunctionto_chat
(raw_prompts: list[str])
experiments/explorations/zero_shot_classification_eval.py:133
↓ 1 callersMethodto_prob
(self)
nl_probes/autointerp_detection_eval/caller.py:216
↓ 1 callersMethodto_prompt_str
(self)
nl_probes/autointerp_detection_eval/detection_basemodels.py:24
↓ 1 callersMethodto_sae_explained
(self, sae_info: SAEInfo)
nl_probes/autointerp_detection_eval/eval_detection_v2.py:367
↓ 1 callersFunctiontrain_model
( cfg: SelfInterpTrainingConfig, training_data: list[TrainingDataPoint], eval_datasets: dict[str,
nl_probes/sft.py:307
↓ 1 callersFunctiontrain_probe
( X_tr_cpu: torch.Tensor, y_tr_cpu: torch.Tensor, X_te_cpu: torch.Tensor, y_te_cpu: torch.Tens
experiments/linear_probe.py:177
↓ 1 callersFunctiontrain_with_sft_only
( sft_train_ds: Dataset, sft_hf_eval_test_ds: Dataset, wandb_sft_project: str, config: EvalCon
nl_probes/trl_training/personaqa_train.py:76
↓ 1 callersFunctiontrain_with_sft_only
( sft_train_ds: Dataset, sft_hf_eval_test_ds: Dataset, wandb_sft_project: str, config: EvalCon
nl_probes/trl_training/taboo_train.py:51
↓ 1 callersFunctionview_tokens
(tokens_L: list[int], tokenizer: AutoTokenizer, offset: int)
nl_probes/dataset_classes/classification.py:156
↓ 1 callersFunctionvisualize_grades_by_type_average
( entries: List[Tuple[str, str, str]], *, config_path: str = CONFIG_PATH, export_dir: str = "e
experiments/final_paper_plots/plot_em_agent_multi_bars.py:376
↓ 1 callersFunctionvisualize_grades_by_type_average
( entries: List[Tuple[str, str, str]], *, config_path: str = CONFIG_PATH, export_dir: str = "e
experiments/final_paper_plots/plot_em_agent.py:364
↓ 1 callersMethodwith_explanation_only
(cls, sae_explanation: SAEExplained)
nl_probes/dataset_classes/sae_training_data.py:81
↓ 1 callersMethodwith_positive_and_negative_examples
(cls, sae_explanation: SAEExplained)
nl_probes/dataset_classes/sae_training_data.py:61
↓ 1 callersMethodwrite_line
(self, key: str, response_json: str)
nl_probes/autointerp_detection_eval/caller.py:514
Method__aenter__
(self)
nl_probes/autointerp_detection_eval/caller.py:426
Method__aexit__
(self, exc_type, exc_value, traceback)
nl_probes/autointerp_detection_eval/caller.py:429
Method__call__
(self, query)
datasets/latentqa_datasets/curate_gpt_data.py:25
Method__getitem__
(self, idx: int)
nl_probes/dataset_classes/misc/latentqa_loader.py:228
Method__init__
(self, d: int, k: int)
experiments/linear_probe.py:169
Method__init__
(self, prompt)
datasets/latentqa_datasets/curate_gpt_data.py:19
Method__init__
( self, d_in: int, d_sae: int, model_name: str, hook_layer: int,
nl_probes/sae.py:146
Method__init__
( self, d_in: int, d_sae: int, k: int, model_name: str, hook_l
nl_probes/sae.py:221
Method__init__
(self, cache_path: Path | str, response_type: type[GenericBaseModel])
nl_probes/autointerp_detection_eval/caller.py:434
Method__init__
(self, cache_path: Path, cache_type: type[GenericBaseModel] = OpenaiResponse)
nl_probes/autointerp_detection_eval/caller.py:527
Method__init__
( self, cache_path: Path | str | CallerCache, api_key: str | None = None, orga
nl_probes/autointerp_detection_eval/caller.py:559
Method__init__
( self, cache_path: Path | str | CallerCache, anthropic_client: anthropic.AsyncAnthrop
nl_probes/autointerp_detection_eval/caller.py:804
Method__init__
(self, clients: Sequence[CallerConfig])
nl_probes/autointerp_detection_eval/caller.py:907
Method__init__
(self, callers: Sequence[Caller])
nl_probes/autointerp_detection_eval/caller.py:956
Method__init__
(self, api_key: str, cache_path: Path | str)
nl_probes/autointerp_detection_eval/caller.py:991
Method__init__
(self, dataset_config: DatasetLoaderConfig, model_kwargs: dict[str, Any] | None = None, model=None)
nl_probes/dataset_classes/classification.py:44
Method__init__
( self, dataset_config: DatasetLoaderConfig, )
nl_probes/dataset_classes/act_dataset_manager.py:51
Method__init__
( self, dataset_config: DatasetLoaderConfig, )
nl_probes/dataset_classes/latentqa_dataset.py:33
Method__init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:52
Method__init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:111
Method__init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:147
Method__init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:196
Method__init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:327
Method__init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:422
Method__init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:462
Method__init__
(self, name: str)
nl_probes/dataset_classes/classification_dataset_manager.py:549
Method__init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:606
Method__init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:724
Method__init__
(self, examples, batch_size, shuffle)
nl_probes/dataset_classes/classification_dataset_manager.py:790
Method__init__
( self, dataset_config: DatasetLoaderConfig, )
nl_probes/dataset_classes/past_lens_dataset.py:35
Method__init__
(self, dataset_config: DatasetLoaderConfig)
nl_probes/dataset_classes/sae_training_data.py:181
Method__init__
(self, dataset_config: DatasetLoaderConfig)
nl_probes/dataset_classes/sae_training_data.py:249
Method__init__
( self, data_groups: Sequence[Dict[str, List[Tuple[str, str, str, str, str, str, str]]]],
nl_probes/dataset_classes/misc/latentqa_loader.py:199
Method__init__
(self)
nl_probes/trl_training/config.py:87
Method__iter__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:812
Method__len__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:809
Method__len__
(self)
nl_probes/dataset_classes/misc/latentqa_loader.py:225
Method__post_init__
Validate configuration.
nl_probes/base_experiment.py:99
Method__post_init__
(self)
nl_probes/dataset_classes/classification_dataset_manager.py:29
Method__str__
(self)
datasets/latentqa_datasets/curate_gpt_data.py:22
Method_check_context_alignment
(cls, values)
nl_probes/utils/dataset_utils.py:57
Function_load_grade_score
(json_path: Path)
experiments/final_paper_plots/plot_agent_adam.py:166
Function_load_grade_score
(json_path: Path)
experiments/final_paper_plots/plot_em_agent_multi_bars.py:166
Function_load_grade_score
(json_path: Path)
experiments/final_paper_plots/plot_em_agent.py:158
Methodadd_messages
(self, messages: Sequence[ChatMessage])
nl_probes/autointerp_detection_eval/caller.py:115
Methodall_assistant_messages
(self)
nl_probes/autointerp_detection_eval/caller.py:101
Methodall_responses
(self)
nl_probes/autointerp_detection_eval/caller.py:145
Functionanalyze_results
(results: list[dict])
nl_probes/utils/eval.py:206
Methodas_text
(self)
nl_probes/autointerp_detection_eval/caller.py:104
Functionassert_no_peft_present
Asserts that no PEFT adapters are present or active on the model. Args: model: The model to check. check_for_active_adapter_
nl_probes/utils/common.py:85
Functioncalculate_accuracy
(record)
experiments/final_paper_plots/plot_ssc_results.py:336
Functioncalculate_zero_shot_baseline
Calculate weighted average accuracy for zero-shot baseline.
experiments/final_paper_plots/plot_classification_eval.py:121
Functioncalculate_zero_shot_baseline
Calculate weighted average accuracy for zero-shot baseline.
experiments/final_paper_plots/plot_classification_eval_all_models.py:160
Methodcall
( self, messages: ChatHistory, config: InferenceConfig, try_number: int = 1,
nl_probes/autointerp_detection_eval/caller.py:609
Methodcall
( self, messages: ChatHistory, config: InferenceConfig, try_number: int = 1,
nl_probes/autointerp_detection_eval/caller.py:835
Methodcall
( self, messages: ChatHistory, config: InferenceConfig, try_number: int = 1,
nl_probes/autointerp_detection_eval/caller.py:925
Methodcall
( self, messages: ChatHistory, config: InferenceConfig, try_number: int = 1,
nl_probes/autointerp_detection_eval/caller.py:963
Methodcall_with_log_probs
( self, messages: ChatHistory, config: InferenceConfig, try_number: int = 1,
nl_probes/autointerp_detection_eval/caller.py:760
Methodcall_with_log_probs
( self, messages: ChatHistory, config: InferenceConfig, try_number: int = 1 )
nl_probes/autointerp_detection_eval/caller.py:894
← previousnext →501–600 of 716, ranked by callers