zjunlp

Build and visualize a citation network from a source/target CSV to identify key papers, communities, and emerging hotspots; use when you have citation pairs and need fast literature review or trend analysis.

Verify, normalize, and enrich a single citation or paper identifier. Use when the user pastes a DOI, URL, arXiv ID, PubMed ID, citation string, or paper title and wants it checked.

Compile LaTeX paper to PDF, fix errors, and verify output. Use when user says \"编译论文\", \"compile paper\", \"build PDF\", \"生成PDF\", or wants to compile LaTeX into a submission-ready PDF.

Automatically scan document reference lists and check against Retraction.

Use only SciAtlas search-papers to take a novice user from zero setup to an evidence-grounded evaluation of a research idea's novelty, feasibility, soundness, and differentiation. Trigger when the user asks whether an idea is worth pursuing or needs literature-backed critique.

Use only SciAtlas search-papers to take a novice user from zero setup to literature-grounded research idea seeds. Trigger when the user wants new research directions, hypotheses, cross-topic combinations, project ideas, or brainstorming grounded in retrieved papers.

Use only SciAtlas search-papers to take a novice user from zero setup to prior-art grounding for a research idea. Trigger when the user gives an idea and asks for similar work, differentiation evidence, related work, motivation support, or literature-grounded refinement.

Use only SciAtlas search-papers to take a novice user from zero setup to an evidence-grounded literature review, related-work summary, survey outline, or reading list. Trigger when the user asks for a literature review, related work section, paper map, or topic overview.

Search, download, create, evaluate, and analyze reusable agent skills via SkillNet — the open skill supply chain for AI agents. Use when: (1) Before any multi-step task — search SkillNet for existing skills first, (2) User says "find a skill", "learn this repo/doc", "turn this into a skill", or mentions skillnet, (3) User provides a GitHub URL, PDF, DOCX, PPT, execution logs, or trajectory — create a skill from it, (4) After completing a complex task with non-obvious solutions — create a skill to preserve learnings, (5) User wants to evaluate skill quality, organize/analyze a local skill library, or orchestrate a supported scene. NOT for: single trivial operations (rename variable, fix typo), or tasks with no reusable knowledge.

Navigates the agent to a target appliance (microwave, stove, fridge, or sinkbasin) needed for object processing. Use when you are holding an object that needs heating, cooling, or cleaning and must move to the correct appliance station. Identifies the required appliance from the task context and executes the movement action.

Prepares a household appliance (microwave, oven, toaster, fridge) for use by ensuring it is in the correct open/closed state. Use when the agent needs to heat, cool, or cook an item and must first open or close the appliance before placing an object inside. Takes an appliance identifier as input and outputs a confirmation that the appliance is ready for the next action.

Cleans a specified object using an appropriate cleaning receptacle (e.g., sinkbasin). Use when a task requires an object to be in a clean state (e.g., "clean potato", "wash apple") before proceeding. Navigates to the cleaning location, performs the clean action, and confirms the object is now clean.

Operates a device or appliance (like a desklamp, microwave, or fridge) to interact with another object. Use when the task requires using a tool on a target item (e.g., "look at laptop under the desklamp", "heat potato with microwave"). Locates both the device and target object, co-locates them, and executes the appropriate use action (toggle, heat, cool, or clean).

Performs an initial scan of the ALFWorld environment to identify all visible objects and receptacles. Use when you first enter an environment and need to build a mental map for task planning. Processes raw observation text into a structured list of entities, categorizing them as objects or receptacles.

Uses a heating appliance (microwave, stoveburner, oven) to apply heat to a specified object. Use when the task requires warming or cooking an item (e.g., "heat some egg", "warm the mug") and a heating appliance is available. Takes the object name and appliance name as input and outputs the object in a heated state, ready for placement at the task's target location.

Use when the agent must collect and track multiple instances of the same object type in ALFWorld (e.g., "put two cellphone in bed"). This skill maintains a count of collected versus needed objects, guides systematic searching through receptacles, and ensures each found object is placed at the target before searching for the next.

Solve questions about which fee IDs apply to a payment merchant, transaction characteristics, or time period in the dabstep dataset. Use this skill for any question asking "which fee IDs apply to X", "what are the applicable fee IDs for merchant Y", "which merchants are affected by fee Z", or any query involving filtering fees.json based on payment characteristics or merchant attributes.

Solve dabstep Average_Fee_Estimation problems. Use this skill for questions asking about average payment processing fees, which card scheme is cheapest/most expensive in an average scenario, and fee calculations based on card scheme, account type, MCC description, credit/debit type, or any combination of these filters.

Use this skill for CSV data analysis tasks that require reading a local CSV file, checking row counts and columns, grouping records, computing rates or aggregates, creating a chart, and writing a short Markdown report.

Solve Applicable_Fee_IDs questions in the dabstep dataset. Use this skill whenever a question asks which fee IDs apply to a merchant, a specific account_type/aci combination, a specific day/month/year, or when reverse-looking up which merchants are affected by a given fee ID.

Solve Average_Fee_Estimation questions in the dabstep dataset. Use this skill when asked about the average fee a card scheme charges for a given transaction value, or which card scheme is cheapest/most expensive in an average scenario. Handles filtering by credit type, account type, and merchant category code (MCC).

Skill for computing average transaction value statistics from payment transaction data in the dabstep dataset. Use this skill when a question asks about average transaction amount/value grouped by a categorical field (e.g., shopper_interaction, issuing_country, acquirer_country, aci), possibly filtered by merchant, card scheme, and/or date range.

Answers questions about the structure, metadata, field definitions, and business rules of the dabstep payment processing dataset. Use this skill when questions ask about: column names, field meanings, fee rule structure, which factors affect fees, fee formula, boolean factor effects on cost, volume/fraud/capture-delay tier meanings, or whether specific fee types exist. Also use for "Not Applicable" determination when a business concept is mentioned but no corresponding fee or data field exists.

Solve dabstep Fee_Delta_and_Impact_Simulation questions: computing fee deltas when a fee's rate changes, and identifying which merchants are affected by fee rule changes. Use when asked about fee impact, delta payments, rate changes, or which merchants would be affected by modifying a fee rule.

Representation and Parameter Analysis interprets and controls a model by directly manipulating its two kinds of internal objects — features (the hidden-state activations produced during a forward pass) and weights (the parameters of the target model). The analysis side relies on standard linear-algebra tools applied straight to these objects: techniques such as PCA, mean-difference, or simple linear classifiers are run over collected features (e.g., from contrastive prompt pairs) or over weight differences between a fine-tuned and a pre-trained checkpoint, in order to extract a small set of meaningful directions. The control side then operates directly through arithmetic on those objects: adding such a direction back into the features steers behavior at inference time, while adding, subtracting, or combining task-level weight differences edits the model in one shot — so that interpretability findings translate into behavioral control without any retraining.

Use this skill for feature-level steering of models — locating the internal feature that drives a target behavior, scoring and selecting it by its effect on the model's output, and directly amplifying or shrinking that feature's activation during generation to control behavior. Applies to features read from the model's own activations or from a Sparse Autoencoder (SAE); the bundled demo scripts happen to use an SAE, but the method does not require one.

Routing entry point for eleven families of mechanistic-interpretability methods that localize *which* internal object (layer, attention head, neuron, SAE feature, weight, or input feature) drives a model's behavior, *how influential* it is, and *what changes* when it is intervened on. Use this skill whenever the question is about a model's internal mechanism rather than its external metrics — for example, claims that a specific component is responsible for a behavior, mechanistic-evidence requests, circuit-discovery tasks, feature-attribution work (SHAP), or concept-level explanations of vision/VL models. The file lays out each family's premise, signal, cost, advantages, limitations, and how to compose them into a cheap-screen → causal-verify pipeline. Loading is hierarchical and mandatory: after picking a family from this routing file you MUST load that family's `SKILL.md`, and after picking a submethod you MUST load that submethod's `SKILL.md` — never act on the previews in this file alone.

Vocabulary Projection methods interpret internal model states by projecting them through the unembedding matrix to obtain a distribution over the vocabulary. The core idea is that the unembedding matrix, which maps the final hidden state to output logits, can serve as a universal decoder for intermediate states throughout the model. By inspecting the top-ranked tokens of the resulting distribution, researchers can directly read off the semantic content encoded in any internal object — such as a residual stream state or an attention head output — in terms of the model's output vocabulary.

Use when main results pass result-to-claim (claim_supported=yes or partial) and ablation studies are needed for paper submission. The external LLM reviewer (via llm-chat MCP) designs ablations from a reviewer's perspective, CC reviews feasibility and implements.

Search, download, and summarize academic papers from arXiv. Use when user says "search arxiv", "download paper", "fetch arxiv", "arxiv search", "get paper pdf", or wants to find and save papers from arXiv to the local paper library.

Workflow 1: Claim-stage pipeline, controlled by two orthogonal axes. BEHAVIOR_SOURCE selects the behavior stage: `given` (default; behavior taken from task.md and assumed to hold — no ideation, no novelty, no M0), `given-validation` (behavior taken from task.md but the experiment plan opens with an M0 phenomenon-validation gate), or `discovery` (mine a NEW behavior via /mechanism-behavior-discovery + full ideation: research-lit → idea-creator → novelty-check → impact-check → research-review → research-refine-pipeline, impact-first ranking; plan opens with M0). MECHANISM selects the mechanism stage: `discovery` (default; system routes — claim loads /mechanism-explore to shape direction) or `given` (user named the mechanism method/family in task.md — capture it and forward as CHOSEN_FAMILY, no routing). resource_fidelity:strict (exact models/datasets, no downscaling) is stamped iff BEHAVIOR_SOURCE=given AND MECHANISM=given (the reproduction combination). Always emits `idea-stage/IDEA_REPORT.md` + `refine-logs/{

Workflow 1.5: Bridge between idea discovery and auto review. Reads EXPERIMENT_PLAN.md, routes mechanism family inline (Phase 1.5), implements experiment code, deploys to GPU, and collects initial results. Use when user says "implement experiments", "experiment", "deploy the plan", or has an experiment plan ready to execute.

Use only the current SciAtlas automated review workflow (`review_pipeline`) to take a novice user from zero setup to a final automated review of a research idea or paper, including setup, registration guidance, workflow configuration, retrieval, artifact reading, and synthesis. Trigger when the user asks whether an idea is worth pursuing, wants automatic review, meta-review, rubric-based critique, novelty/feasibility/soundness assessment, or literature-backed evaluation.

Use only the current SciAtlas multi-step idea-generation workflow (`sciatlas_idea_gen`) to take a novice user from zero setup to final literature-grounded research idea seeds, including setup, registration guidance, workflow configuration, KG retrieval, artifact reading, novelty checking, and synthesis. Trigger when the user wants new research directions, hypotheses, cross-topic combinations, project ideas, or brainstorming grounded in SciAtlas KG retrieval.

Use only SciAtlas search-papers to take a novice user from zero setup to final prior-art grounding for a research idea, including setup, registration guidance, configuration, retrieval, artifact reading, and synthesis. Trigger when the user gives an idea and asks for similar work, differentiation evidence, related work, motivation support, or literature-grounded refinement.

Use only the current SciAtlas literature-review workflow (`literature_review_pipeline`) to take a novice user from zero setup to a final evidence-grounded survey outline, paper map, or literature review, including setup, registration guidance, workflow configuration, SciAtlas backend retrieval, artifact reading, and synthesis. Trigger when the user asks for a literature review, related work section, paper map, survey outline, reading path, or topic overview grounded in the provided SciAtlas paper backend.

Use only SciAtlas search-papers to take a novice user from zero setup to a final small evidence-backed paper search result, including setup, registration guidance, configuration, retrieval, artifact reading, and summary. Trigger when the user wants a quick paper check, first-pass literature evidence, a retrieval smoke test, or help deciding which deeper SciAtlas agent skill to use next.

Use only SciAtlas search-papers to take a novice user from zero setup to a final evidence-grounded researcher profile from retrieved papers, including setup, registration guidance, configuration, retrieval, artifact reading, and synthesis. Trigger when the user asks for a researcher's background, representative works, topic trajectory, or author-centered literature overview.

Use only SciAtlas search-papers to take a novice user from zero setup to a final timeline-oriented research trend report, including setup, registration guidance, configuration, retrieval, artifact reading, and synthesis. Trigger when the user asks for topic history, field evolution, recent trends, representative papers over time, or emerging directions.

An expert Dify workflow builder that designs and generates Dify-importable workflows through multi-round conversation. Produces structured JSON (nodes, edges, variable references) convertible to Dify YAML, with support for incremental add/modify/remove operations on existing workflows.

A design-only workflow designer for the Dify and Coze platforms. Through multi-round conversation, it produces a structured workflow JSON (nodes, edges, variable references) as text output. The skill itself produces only text and never runs any scripts.

Validation-gated skill evolution from agent failure trajectories (step-level t★, Validator Δ>0). Use when: (1) Agent failed on a task and you have execution trajectories or logs, (2) User asks to evolve, adapt, or improve SKILL.md from failures, (3) Benchmark run produced trajectories under .skill-adaptor/artifacts/, (4) User wants training-free skill bank growth with held-out validation — NOT one-shot log-to-skill. Prefer SkillNet ($skillnet) search first; use this skill when failures need verified adaptation. NOT for: packaging a chat log without re-execution (use skillnet create instead).

Run SkillAdaptor to evolve agent skills from task failures. Use when improving SKILL.md files from trajectories or after benchmark runs.