Exam Bundle

Notebook grounding depends on the actual selected reference set and permissions. For the stated situation—A shared research workspace displays document references that some participants are not permitted to access—this result does not establish the required evidence: “Participants only receive content they are authorized to access and access requests reach the right owner”.

Chat history navigation supports revisiting an exchange; creating a new chat is not reliable retrieval of prior text. For the stated situation—A shared research workspace displays document references that some participants are not permitted to access—this result does not establish the required evidence: “Participants only receive content they are authorized to access and access requests reach the right owner”.

Prompts can request evidence and uncertainty, but the resulting citations still require checking. For the stated situation—A shared research workspace displays document references that some participants are not permitted to access—this result does not establish the required evidence: “Participants only receive content they are authorized to access and access requests reach the right owner”.

Conversation turns share context, so separating unrelated work reduces accidental cross-topic confusion. That does not meet the situation described—A researcher wants a notebook to answer only from uploaded regulatory papers, not unrelated email. The more relevant action is “Explicitly select the Notebook's reference set and ask for citations from those references”.

Chat renaming is an organization step; it does not change source permissions or the answer's accuracy. That does not meet the situation described—A researcher wants a notebook to answer only from uploaded regulatory papers, not unrelated email. The more relevant action is “Explicitly select the Notebook's reference set and ask for citations from those references”.

Change summaries require an explicit baseline; using only the latest document cannot prove what changed. That does not meet the situation described—A researcher wants a notebook to answer only from uploaded regulatory papers, not unrelated email. The more relevant action is “Explicitly select the Notebook's reference set and ask for citations from those references”.

Generated recollections can diverge from historical text; chat history is the proper source for prior wording. For the stated situation—A researcher wants a notebook to answer only from uploaded regulatory papers, not unrelated email—this result does not establish the required evidence: “Answers use the chosen materials and identify unsupported questions as gaps”.

Chat history navigation supports revisiting an exchange; creating a new chat is not reliable retrieval of prior text. For the stated situation—A researcher wants a notebook to answer only from uploaded regulatory papers, not unrelated email—this result does not establish the required evidence: “Answers use the chosen materials and identify unsupported questions as gaps”.

A prompt should distinguish content generation from permitted external action, particularly for sensitive messages. For the stated situation—A researcher wants a notebook to answer only from uploaded regulatory papers, not unrelated email—this result does not establish the required evidence: “Answers use the chosen materials and identify unsupported questions as gaps”.

Conversation turns share context, so separating unrelated work reduces accidental cross-topic confusion. That does not meet the situation described—A training coordinator needs a controlled reference collection for a policy Q&A exercise. The more relevant action is “Explicitly select the Notebook's reference set and ask for citations from those references”.

A chat response may contain confidential facts even if the chat interface itself is private to the requester. That does not meet the situation described—A training coordinator needs a controlled reference collection for a policy Q&A exercise. The more relevant action is “Explicitly select the Notebook's reference set and ask for citations from those references”.

Iterative prompting uses feedback and follow-up context to improve a draft rather than starting from scratch blindly. That does not meet the situation described—A training coordinator needs a controlled reference collection for a policy Q&A exercise. The more relevant action is “Explicitly select the Notebook's reference set and ask for citations from those references”.

Generated recollections can diverge from historical text; chat history is the proper source for prior wording. For the stated situation—A training coordinator needs a controlled reference collection for a policy Q&A exercise—this result does not establish the required evidence: “Answers use the chosen materials and identify unsupported questions as gaps”.

Chat renaming is an organization step; it does not change source permissions or the answer's accuracy. For the stated situation—A training coordinator needs a controlled reference collection for a policy Q&A exercise—this result does not establish the required evidence: “Answers use the chosen materials and identify unsupported questions as gaps”.

Constrained decision prompts make recommendations accountable to measurable user priorities. For the stated situation—A training coordinator needs a controlled reference collection for a policy Q&A exercise—this result does not establish the required evidence: “Answers use the chosen materials and identify unsupported questions as gaps”.

A curated reference scope improves relevance and prevents unrelated workplace data from being treated as evidence. That does not meet the situation described—A user has a long Copilot chat and wants to work on a new unrelated confidential client matter. The more relevant action is “Start a separate chat for the unrelated task with its own clearly stated authorized context”.

Combining curated references with clearly defined questions improves reproducibility of knowledge work. That does not meet the situation described—A user has a long Copilot chat and wants to work on a new unrelated confidential client matter. The more relevant action is “Start a separate chat for the unrelated task with its own clearly stated authorized context”.

Change summaries require an explicit baseline; using only the latest document cannot prove what changed. That does not meet the situation described—A user has a long Copilot chat and wants to work on a new unrelated confidential client matter. The more relevant action is “Start a separate chat for the unrelated task with its own clearly stated authorized context”.

Chat history navigation supports revisiting an exchange; creating a new chat is not reliable retrieval of prior text. For the stated situation—A user has a long Copilot chat and wants to work on a new unrelated confidential client matter—this result does not establish the required evidence: “The new result uses the second project's sources without old-chat assumptions leaking into it”.

Reference-grounded responses should not pretend the selected corpus contains information that it does not. For the stated situation—A user has a long Copilot chat and wants to work on a new unrelated confidential client matter—this result does not establish the required evidence: “The new result uses the second project's sources without old-chat assumptions leaking into it”.

Ambiguous objectives should be clarified before an apparently authoritative recommendation is produced. For the stated situation—A user has a long Copilot chat and wants to work on a new unrelated confidential client matter—this result does not establish the required evidence: “The new result uses the second project's sources without old-chat assumptions leaking into it”.

A curated reference scope improves relevance and prevents unrelated workplace data from being treated as evidence. That does not meet the situation described—A consultant notices that follow-up prompts inherit previous project assumptions in the same conversation. The more relevant action is “Start a separate chat for the unrelated task with its own clearly stated authorized context”.

Combining curated references with clearly defined questions improves reproducibility of knowledge work. That does not meet the situation described—A consultant notices that follow-up prompts inherit previous project assumptions in the same conversation. The more relevant action is “Start a separate chat for the unrelated task with its own clearly stated authorized context”.

Prompts can request evidence and uncertainty, but the resulting citations still require checking. That does not meet the situation described—A consultant notices that follow-up prompts inherit previous project assumptions in the same conversation. The more relevant action is “Start a separate chat for the unrelated task with its own clearly stated authorized context”.

Chat history navigation supports revisiting an exchange; creating a new chat is not reliable retrieval of prior text. For the stated situation—A consultant notices that follow-up prompts inherit previous project assumptions in the same conversation—this result does not establish the required evidence: “The new result uses the second project's sources without old-chat assumptions leaking into it”.

Explicit identity and source disambiguation reduces confusion between similarly named entities in chat context. For the stated situation—A consultant notices that follow-up prompts inherit previous project assumptions in the same conversation—this result does not establish the required evidence: “The new result uses the second project's sources without old-chat assumptions leaking into it”.

Model choice affects capability and efficiency, but no model removes the need to review sensitive outputs. For the stated situation—A consultant notices that follow-up prompts inherit previous project assumptions in the same conversation—this result does not establish the required evidence: “The new result uses the second project's sources without old-chat assumptions leaking into it”.

Explicit identity and source disambiguation reduces confusion between similarly named entities in chat context. That does not meet the situation described—A project officer wants the exact wording of a previous draft rather than a newly generated paraphrase. The more relevant action is “Reopen the existing chat and inspect the original prompt and reply rather than asking the model to reconstruct them”.

A chat response may contain confidential facts even if the chat interface itself is private to the requester. That does not meet the situation described—A project officer wants the exact wording of a previous draft rather than a newly generated paraphrase. The more relevant action is “Reopen the existing chat and inspect the original prompt and reply rather than asking the model to reconstruct them”.

Model choice affects capability and efficiency, but no model removes the need to review sensitive outputs. That does not meet the situation described—A project officer wants the exact wording of a previous draft rather than a newly generated paraphrase. The more relevant action is “Reopen the existing chat and inspect the original prompt and reply rather than asking the model to reconstruct them”.

A curated reference scope improves relevance and prevents unrelated workplace data from being treated as evidence. For the stated situation—A project officer wants the exact wording of a previous draft rather than a newly generated paraphrase—this result does not establish the required evidence: “The reviewer sees the actual stored exchange, with no invented recollection”.

Notebooks collect and scope reference materials for sustained work rather than relying on diffuse chat history. For the stated situation—A project officer wants the exact wording of a previous draft rather than a newly generated paraphrase—this result does not establish the required evidence: “The reviewer sees the actual stored exchange, with no invented recollection”.

Saving a reusable prompt preserves the instruction template but does not itself execute future runs. For the stated situation—A project officer wants the exact wording of a previous draft rather than a newly generated paraphrase—this result does not establish the required evidence: “The reviewer sees the actual stored exchange, with no invented recollection”.