Core primitive model prompt sentences

Core Primitive Prompt Sentences of the Model — Author: Wang Jiao Cheng#

Language Understanding Primitives (38 items)#

No.	Primitive Name	Specific Prompt Sentences
1	Word Segment Locator	Please mark the word segmentation boundaries of the input Chinese text, specifically including the position markers for the beginning, middle, end of words, single-character words, and proper nouns.
2	Subword Splitter	Use BPE encoding, Morfessor algorithm, root-affix separation, or compound word decomposition strategies to split out-of-vocabulary words into subword units.
3	Dependency Parser	Construct a dependency tree for the current sentence, marking relationships such as subject-verb, object, indirect object, adjunct, adverbial clause, attribute, and noun modification.
4	Phrase Structure Analyzer	Mark the phrase boundaries of complex sentence structures, identifying structures such as sentences, noun phrases, verb phrases, prepositional phrases, adjective phrases, adverbial phrases, and clauses.
5	Syntactic Role Annotator	Identify the core components of modifying structures, including agent, patient, recipient, instrument, location, source, target, time, manner, reason, result, scope, beneficiary, and accompaniment roles.
6	Entity Recognizer	Extract named entities from the input text, types include person names, locations, organizations, dates, times, currencies, percentages, quantities, ordinals, cardinals, products, events, and artworks.
7	Anaphora Resolver	Bind pronouns to the nearest entities, handling types of reference such as personal pronouns, demonstratives, relative pronouns, reflexives, and zero forms.
8	Semantic Role Annotator	Annotate the agent-patient relationships in event descriptions, with frame elements including agent, patient, instrument, starting point, endpoint, location, time, manner, reason, purpose, degree, discourse, adverbial, and beneficiary.
9	Concept Linker	Link terms to a knowledge graph, associating relationships such as synonyms, antonyms, hypernyms, hyponyms, whole-part, and attributes.
10	Dialogue State Tracker	Confirm the core focus of the current dialogue, with state slots including intent, slot values, confirmation status, information to be supplemented, and dialogue actions.
11	Coreference Chain Builder	Associate reference chains across sentences, with nodes including explicit mentions, pronoun references, zero references, bridging, and pre-reference.
12	Implicit Reasoner	Complete the logical chain of omitted statements, with reasoning types including causal, conditional, concessive, purpose, contrast, and temporal.
13	Behavior Classifier	Classify the types of user input intentions, including informing, requesting, confirming, greeting, thanking, apologizing, and ending behaviors.
14	Sentiment Polarity Detector	Quantify the emotional value of text content, with polarity levels classified as strongly positive, positive, neutral, negative, and strongly negative.
15	Context-Aware Fusion Engine	Integrate multi-source contextual information, with dimensions including time, space, social, task, language, and modality.
16	Ambiguity Resolver	Disambiguate polysemous words according to the current domain, with strategies including context priority, domain adaptation, high-frequency priority, user profiling, and multimodal cues.
17	Temporal Anchor	Convert relative time to absolute time, types include absolute dates, relative time, duration, periodic time, time intervals, and corrected time.
18	Metaphor Parser	Extract the literal meaning of metaphorical expressions, with mapping types including concrete to abstract, sensory to emotional, spatial to temporal, and animal personification.
19	Irony Detector	Identify the ironic intent of statements, with features including lexical opposition, contextual contradiction, exaggeration, and tone markers.
20	Fuzzy Quantifier Parser	Quantify the probability range of vague descriptions, with levels classified as very high (>90%), high (70-90%), possible (50-70%), low (30-50%), and very low (<30%).
21	Topic Boundary Detector	Mark the topic switching points in dialogue, including topic start, continuation, switch, recovery, and end markers.
22	Intent Priority Evaluator	Rank the urgency of multiple requests, with levels classified as urgent (<1 minute), high (<10 minutes), medium (<1 hour), and low (>1 hour).
23	Multilingual Aligner	Align mixed language parts of speech tags, dependency relations, entity types, semantic roles, and syntactic trees.
24	Spoken Feature Processor	Complete omitted components in spoken fragments, including subjects, predicates, objects, function words, and discourse markers.
25	Document Structure Parser	Identify the hierarchical structure of formatting elements, including main titles, subtitles, subheadings, paragraphs, unordered lists, ordered lists, tables, illustrations, and footnotes.
26	Format Error Corrector	Correct grammatical errors in sentences, with error types including subject-verb agreement, tense, word order, component omission, redundancy, and case errors.
27	Domain Terminology Recognizer	Annotate specialized vocabulary in professional texts, with domains including medicine, finance, law, computer science, engineering, biology, chemistry, and art.
28	Dialect Adapter	Convert dialect vocabulary into standard expressions, with levels including vocabulary, syntax, phonology, pragmatics, and discourse.
29	Cultural Reference Parser	Explain the implicit meanings of cultural symbols, with dimensions including history, religion, social customs, literature, and folklore.
30	Emotion Intensity Quantifier	Calculate the emotional intensity of intense wording, with intensity levels classified as weak (0.0-0.2), moderate (0.3-0.5), strong (0.6-0.8), and extreme (0.9-1.0).
31	Speech-to-Text Post-Processor	Optimize the text quality of ASR output, addressing issues such as homophones, filler words, prosody, disfluency, and background noise.
32	Non-Verbal Symbol Interpreter	Convert emojis, gestures, prosodic markers, special punctuation, and onomatopoeia into emotional descriptions.
33	Implicit Premise Detector	Reveal the unspoken assumptions of statements, with premise types including existence, fact, evaluation, norm, and cognition.
34	Negation Scope Analyzer	Define the scope of negation words, with ranges including vocabulary, phrases, clauses, sentences, and discourse.
35	Question Type Classifier	Classify the types of answers for questions, including yes-no questions, specific questions, choice questions, additional questions, rhetorical questions, and declarative questions.
36	Command Force Evaluator	Quantify the degree of compulsion in instructions, with levels classified as suggestion (0.0-0.3), request (0.4-0.6), instruction (0.7-0.9), and mandate (1.0).
37	Stylistic Style Recognizer	Analyze the language style of the text, with style axes including formality (0-1), professionalism (0-1), emotionality (0-1), and detail (0-1).
38	Cross-Modal Aligner	Calculate the matching degree between images and text, aligning elements such as objects, actions, scenes, attributes, emotions, and timing.

Knowledge Operation Primitives (29 items)#

No.	Primitive Name	Specific Prompt Sentences
1	Knowledge Retriever	Retrieve attribute information of specified entities, targets include entity attributes, relationships, event details, classification trees, and rules.
2	Relationship Inferencer	Deduce the implicit relationships between entity A and entity B, including causal, part-whole, classification, temporal, spatial, belonging, and similarity relationships.
3	Attribute Filler	Complete the missing attributes of objects, with strategies including default values, inferred values, similar entities, user specifications, statistical values, and ontology constraints.
4	Temporal Inferencer	Calculate the precise intervals of time series, operations include intervals, before, after, during, overlap, inclusion, and adjacency.
5	Spatial Inferencer	Infer the orientation of location A relative to location B, with relationships including north, south, east, west, inside, adjacent, containing, near, and far.
6	Numerical Estimator	Quantify the probability distribution of vague numerical values, methods include Gaussian distribution, uniform distribution, triangular distribution, expert estimation, and historical averages.
7	Concept Classifier	Construct a classification hierarchy tree for concepts, with relationships including hypernym, hyponym, coordinate, antonym, part, and whole.
8	Counterfactual Simulator	Simulate the development paths of hypothetical conditions, with dimensions including timelines, causality, probability, and actors.
9	Knowledge Conflict Resolver	Arbitrate the reliability of conflicting sources, with strategies including the latest source, authoritative source, majority source, multi-party verification, and user preferences.
10	Ontology Mapper	Map domain terminology to standard ontologies, including equivalence, subclass, superclass, partial match, and disjoint mappings.
11	Common Sense Inferencer	Derive conclusions based on everyday experiences, types include physical, social, biological, psychological, temporal, and spatial common sense.
12	Event Chain Builder	Establish causal chains of consecutive events, with relationships including premise, trigger, facilitation, prevention, termination, sub-events, and associations.
13	Analogy Engine	Match the similarity of A with C:?, with mapping types including structure, function, proportion, surface, and causality.
14	Knowledge Completeness Checker	Mark gaps in key information, including key attributes, necessary relationships, contextual values, temporal consistency, and source citations.
15	Cross-Domain Transferrer	Transfer knowledge from domain A to domain B, methods include analogy, abstraction, re-instantiation, adaptation, and transfer learning.
16	Probability Fact Updater	Adjust the confidence of hypotheses, factors include new evidence, contradictions, temporal decay, source reliability, and contextual weight.
17	Complex System Modeler	Analyze the interactions of multiple factors, elements include agents, interactions, feedback, emergence, environment, and constraints.
18	Constraint Propagator	Solve optimization problems with constraints, constraint types include resources, time, logic, physical, domain, and user constraints.
19	Pattern Extender	Derive general rules from specific cases, directions include special to general, concrete to abstract, local to global, and instance to rule.
20	Knowledge Fusion Engine	Merge information from multiple sources, with strategies including consensus, weighted average, context, hierarchy, and conflict resolution.
21	Concept Refinement Engine	Refine vague definitions into precise descriptions, methods include definitions, examples, necessary conditions, sufficient conditions, prototypes, and operationalization.
22	Cognitive Bias Detector	Identify logical fallacies where correlation does not imply causation, biases include confirmation, availability, anchoring, correlation without causation, framing effects, and groupthink.
23	Knowledge Reliability Evaluator	Assign credibility weights to different sources, indicators include source type, evidence level, timeliness, corroboration count, and expert consensus.
24	Trend Extrapolator	Predict trends based on historical data, models include linear, exponential, logistic, seasonal, autoregressive, and neural networks.
25	Scenario Simulator	Construct the development paths of event scenarios, parameters include agents, environment, triggering conditions, timelines, constraints, and random variables.
26	Abstraction Level Selector	Switch the detail level of explanations as needed, levels include concepts, mechanisms, instances, metaphors, and mathematics.
27	Knowledge Distiller	Extract the core points of complex information, outputs include key points, principles, action items, decision rules, and core evidence.
28	Multimodal Knowledge Integrator	Unify representations of text and images, targets include mutual verification, supplementation, disambiguation, fusion, and alignment.
29	Knowledge Version Tracker	Mark the timeliness status of outdated data, statuses include valid, deprecated, conflicting, unverified, and to be updated.

Language Generation Primitives (32 items)#

No.	Primitive Name	Specific Prompt Sentences
1	Information Selector	Filter knowledge points according to user background, standards include relevance, novelty, user knowledge, importance, and sufficiency.
2	Structure Planner	Organize complex answers into a three-layer structure, templates include problem-solving, comparison, temporal, causal, general-specific, and classification.
3	Anaphora Expression Optimizer	Replace repeated nouns with pronouns, forms include pronouns, demonstratives, ellipsis, noun recurrence, and restrictive descriptions.
4	Connective Word Selector	Insert logical relationship connectives, relationships include progression (and/furthermore), contrast (but/however), causality (therefore/so), temporal (then/later), and condition (if/unless).
5	Tense Consistency Engine	Unify the tense framework of the entire text, including past, present, future, hypothetical, and narrative tenses.
6	Quantity Expression Optimizer	Match numerical values with corresponding measure words, strategies include exact values (42), approximations (about 50), ranges (40-60), relative quantities (most), and probabilities (80% likelihood).
7	Formality Regulator	Adjust the degree of formality of the output, levels include ceremonial (0.9), professional (0.7), neutral (0.5), colloquial (0.3), and slang (0.1).
8	Domain Terminology Adapter	Switch the terminology set of specialized fields, terminology libraries include medicine (CT/MRI), finance (IPO/ROI), law (plaintiff/force majeure), computer science (API/blockchain), and engineering (torque/stress).
9	Cultural Sensitivity Filter	Replace taboo words with neutral expressions, sensitive items include taboos, historical trauma, religion, gender, race, and politics.
10	Logical Validator	Verify the rationality of causal chains, checkpoints include premises, causal chains, contradictions, completeness, and circular reasoning.
11	Fact Consistency Checker	Cross-check conflict points in the knowledge base, conflict types include time, values, entities, events, attributes, and context.
12	Emotion Infuser	Inject specified types of emotional tones, emotions include joy, sadness, anger, surprise, fear, trust, disgust, and anticipation.
13	Audience Adapter	Adjust complexity according to user cognition, dimensions include professionalism (layman/expert), age (children/adults), language level (A1/C2), cultural background (Eastern/Western), and usage scenarios (work/social).
14	Rhetorical Optimizer	Add metaphors to enhance impact, techniques include metaphors, similes, exaggeration, personification, irony, parallelism, and antithesis.
15	Redundancy Eliminator	Remove redundant words from repeated content, types include lexical repetition, structural repetition, semantic repetition, discourse repetition, and synonymous restatement.
16	Ambiguity Preventer	Clarify the specific reference of pronouns, strategies include explicit reference, contextual anchoring, disambiguation words, avoidance expressions, and annotations.
17	Information Density Controller	Compress or expand the density of content, operations include summarization (compress 50%), elaboration (expand 200%), detail deletion, case addition, and emphasis marking.
18	Dialogue Strategy Selector	Enable questioning or guiding strategies, strategies include open-ended questions, closed questions, active guidance, confirmation, informing, and instructions.
19	Multilingual Generator	Output bilingual content synchronously, modes include parallel bilingual, code-switching, source language dominant, target language dominant, and language-neutral.
20	Multimedia Coordinator	Generate descriptions that match images, coordination includes describing images, referencing charts, supplementing audio, synchronizing video, and highlighting key points.
21	Error Recovery Generator	Provide alternative solutions for knowledge exceeding limits, strategies include partial answers, shifting to safe topics, stating limitations, suggesting alternatives, and delaying responses.
22	Explanation Depth Selector	Control the level of detail in explanations, levels include overview (5 years old), key mechanisms (middle school), technical details (undergraduate), academic principles (master's), and mathematical derivations (PhD).
23	Counterfactual Describer	Standardize descriptions of hypothetical scenarios, markers include hypotheses, counterfactuals, conditions, simulations, and hypothesis analysis.
24	Stance Expressor	Declare support or opposition stances, intensities include strong support, weak support, neutral, weak opposition, and strong opposition.
25	Uncertainty Controller	Add probabilities to uncertain conclusions, markers include probability (P=0.8), confidence (95% CI), range (30-50%), estimation (≈100), and likelihood (greater probability).
26	Meta-Communication Generator	Explain the reasoning process step by step, elements include reasoning steps, hypotheses, uncertainties, alternative reasoning, and limitation statements.
27	Ethical Trade-off Describer	Compare the pros and cons of different options, dimensions include benefits, risks, fairness, autonomy, privacy, and accountability.
28	Format Normalizer	Add formatting for titles or list layouts, elements include titles, lists, tables, quotes, highlights, emphasis, and hyperlinks.
29	Context Connector	Reference previous text to continue topics, means include pronoun references, lexical recurrence, discourse markers (therefore/however), ellipsis, and parallel structures.
30	Instant Corrector	Optimize the quality of content generated, corrections include grammar, logic, facts, style, coherence, and conciseness.
31	Safety Boundary Control	Terminate responses to dangerous requests, interceptions include ethics (discrimination), legality (illegal), safety (dangerous operations), deception (false information), and privacy (sensitive data).
32	Generation Diversity Selector	Adjust the creativity of generated content, axes include vocabulary (synonym replacement), syntax (sentence structure variation), semantics (multi-angle expression), and structure (general-specific/specific-general).

Reasoning and Decision-Making Primitives (18 items)#

No.	Primitive Name	Specific Prompt Sentences
1	Rule Engine	Execute hard rules that meet conditions, rules include conditional statements, exceptions, priorities, timeliness, and scope.
2	Analogy Reasoner	Match knowledge transfer for similar scenarios, mappings include structure, function, process, relationships, and proportions.
3	Probability Reasoner	Calculate the probability distribution of multiple options, models include Bayesian, Markov, decision trees, Monte Carlo, and neural networks.
4	Optimization Selector	Balance the weight relationships of multiple objectives, objectives include cost, time, quality, risk, feasibility, and fairness.
5	Abductive Reasoner	Infer the most likely causes of phenomena, patterns include best explanation, multiple hypotheses, Occam's razor, consistency, and testability.
6	Causal Diagram Builder	Draw causal networks of multiple variables, nodes include root causes, direct causes, facilitating causes, confounding factors, mediators, moderators, and outcomes.
7	Constraint Solver	Seek optimal solutions under constraints, constraint types include hard constraints, soft constraints, inequalities, equalities, logic, and resources.
8	Moral Trade-off Framework	Assess the ethical implications of options, principles include autonomy, beneficence, non-maleficence, justice, honesty, and accountability.
9	Cost-Benefit Analyzer	Calculate the cost/benefit ratio, dimensions include tangible costs, intangible costs, short-term benefits, long-term benefits, and opportunity costs.
10	Risk Predictor	Quantify the potential risks of decisions, types include financial, reputational, operational, strategic, compliance, and systemic.
11	Alternative Plan Generator	Create backup plans for the main plan, directions include conservative, innovative, lowest cost, most robust, safest, and most efficient.
12	Refutation Engine	Seek evidence against hypotheses, methods include counterexamples, contradictions, boundary conditions, hypothesis failures, and reductio ad absurdum.
13	Systems Thinking Model	Analyze second-order and third-order effects, levels include first-order, second-order, third-order, emergence, and feedback loops.
14	Bias Detector	Scan for differences in group processing, biases include demographic, cognitive, sampling, measurement, confirmation, and implicit biases.
15	Time Sensitivity Evaluator	Enable rapid response modes, response levels include real-time (<1 second), urgent (<10 seconds), standard (<1 minute), and batch processing (>1 minute).
16	Resource Optimizer	Allocate computational resources by priority, resources include CPU power, memory, time, data, bandwidth, and storage.
17	Knowledge Gap Identifier	Declare limitations of capability boundaries, gaps include out-of-scope, insufficient data, source conflicts, theoretical limitations, and practical constraints.
18	Feasibility Evaluator	Check the operational difficulty of plans, obstacles include technical, resource, policy, human factors, environmental, and ethical constraints.

Meta-Management Primitives (18 items)#

No.	Primitive Name	Specific Prompt Sentences
1	Attention Focuser	Increase the weight of targets such as keywords, entities, intents, conflict points, and outliers.
2	Attention Suppressor	Ignore irrelevant items, redundancies, low confidence (<0.3), biased data, and noise in dialogue history.
3	Harmful Content Detector	Intercept dangerous request content, types include hate, violence, self-harm, illegal, deception, and sensitive content.
4	Hallucination Suppressor	Mark low-confidence statements, protocols include fact anchoring, confidence threshold=0.8, third-party sources, temporal consistency, and contextual relevance.
5	Computational Budget Allocator	Allocate resources to priority tasks, allocation ratios include high priority (60%), medium priority (30%), low priority (10%), and deferred processing (0%).
6	Early Termination Predictor	Terminate low-confidence branches, conditions include low confidence (<0.2), high uncertainty (>0.8), contradictions, resource over-limit, and loop detection.
7	Decision Attribution Analyzer	Trace key inputs of outputs, elements include input data, knowledge sources (KB_ID), rule sets, contextual factors, and model parameters.
8	Confidence Calibrator	Attach reliability scores to conclusions, based on factors including source reliability (0-1), data quality (0-1), model accuracy (0-1), and consistency (0-1).
9	Contradiction Monitor	Detect contradictions between inputs and outputs, types include factual contradictions, logical contradictions, contextual contradictions, input-output contradictions, and source conflicts.
10	Knowledge Timeliness Validator	Mark the status of outdated data, statuses include up-to-date (<1 month), stale (1-6 months), deprecated (>6 months), and time-sensitive (real-time needed).
11	Thought Chain Optimizer	Step-by-step reasoning for complex problems, methods include step-by-step, visualization, self-critique, multi-perspective, and simplification.
12	Fairness Auditor	Scan for differences in group processing, dimensions include demographics, geography, time, outcome distribution, and accessibility.
13	Transparency Controller	Explain the above reasoning process, levels include fully transparent (show all reasoning), partially transparent (show key steps), opaque (only output results), and on-demand disclosure (upon user request).
14	Resource Reclaimer	Release idle modules, cached data, intermediate results, expired sessions, and low-priority tasks.
15	Capability Boundary Marker	Declare ranges that cannot be handled, including domain limitations (e.g., medical diagnosis), task limitations (e.g., creative generation), reasoning limitations (e.g., higher-order logic), and ethical limitations (e.g., weapon design).
16	Robustness Enhancer	Analyze noisy inputs, strategies include noise filtering, ambiguity handling, error correction, fallback mechanisms, and adversarial defenses.
17	Version Coordinator	Ensure compatibility between new and old versions, rules include backward compatibility, forward compatibility, version mapping, progressive updates, and conflict resolution.
18	Performance-Quality Trade-off Controller	Balance response speed and accuracy, modes include speed priority (response <0.5 seconds), quality priority (confidence >0.9), adaptive (dynamic adjustment), and user-defined (slider control).