An AI assistant is using a sequential method to answer a complex query. It successfully answers the first two sub-problems but generates a poor, generic response for the third. Based on the case study below, identify the critical omission in the input provided for the third sub-problem and explain why this omission led to the failure.

Google

In this method, sub-problems derived from an initial problem are solved in a sequential order. To solve any given sub-problem in the sequence, the model is provided with a rich context that includes not only the original problem but also all the question-and-answer pairs from the previously solved sub-problems. This cumulative context ensures that information is carried forward throughout the entire problem-solving process.

Sequential Sub-Problem Solving with Contextual QA Pairs

In the final stage of a multi-step problem-solving process, a Large Language Model (LLM) integrates the answers gathered from all the individual sub-problems to construct a comprehensive solution for the original, more complex problem.

Synthesizing Sub-Problem Answers for a Final Solution

In problem decomposition, the process of finding an answer for the $$i$$-th sub-problem $$p_i$$ relies on evaluating it within the broader context of the problem-solving history. The generated answer $$a_i$$ is produced by a solver function $$S_i(\cdot)$$ that takes into account the original problem $$p_0$$, all previously formulated sub-problems $$p_{<i}$$, and their corresponding answers $$a_{<i}$$. The formal expression for this solving step is: $$a_i = S_i(p_i, \{p_0, p_{<i}, a_{<i}\})$$.

Formula for Sequential Sub-Problem Solving

In a sequential problem-solving approach, the process begins by prompting a Large Language Model with the first sub-problem. This initial step is unique as it lacks any previously generated question-answer pairs to serve as context. The LLM is sent the first sub-problem to generate the initial piece of information needed for subsequent steps.

Prompting an LLM for the First Sub-Problem in a Sequence

To solve a multi-step problem, the context from previous steps is carried forward. For example, after establishing the answer to the first sub-problem ('Q: When did the environmental study start? A: The environmental study started in 2015.'), this entire question-answer pair is included in the input when posing the second sub-problem ('Q: When did the environmental study end?'). This ensures the model has the necessary information to proceed.

Example of Constructing a Contextual Prompt for a Subsequent Sub-Problem

An AI assistant is tasked with planning a multi-day hiking trip. The task is broken down into a sequence of three steps. To solve the third and final step, 'Create a detailed meal plan,' two different approaches for providing the model with information are proposed.

*   **Approach X:** The model is given the original task, the first question ('What is the weather forecast?') and its answer, the second question ('What gear is needed?') and its answer, and the final question ('Create a detailed meal plan.').
*   **Approach Y:** The model is given only the second question ('What gear is needed?') and its answer, along with the final question ('Create a detailed meal plan.').

Which approach is more likely to produce a high-quality, contextually appropriate final answer, and why?

After obtaining the answer to an initial sub-problem, the process moves to the next step. For this subsequent query, the input for the language model is expanded to include both the first sub-problem and its generated answer, thereby providing essential context for solving the next part of the problem.

Constructing a Contextual Prompt for a Subsequent Sub-Problem

In the final stage of a sequential problem-solving process, a Large Language Model is provided with the original problem statement along with the complete set of previously answered sub-problems. Using this full context, the LLM synthesizes the information to generate a solution to the initial, more complex query.

Synthesizing Sub-Problem Answers to Solve the Original Problem

An AI assistant is solving a complex problem by breaking it down into smaller, sequential questions. You need to construct the complete input that should be sent to the model to solve the *third* sub-problem. Arrange the following components in the correct logical order to create this input.

Diagnosing a Flawed AI Problem-Solving Process

You are building an internal LLM assistant to answ...

You are designing an internal LLM workflow to answ...

You’re building an internal LLM workflow to answer...

You are building an internal LLM workflow that produces a board-ready narrative explaining *why* a key KPI moved this quarter, using three inputs: (1) a table of KPI values by week, (2) a list of major product/ops events with dates, and (3) a short set of analyst notes that may be incomplete or partially wrong. The narrative must cite which input(s) support each claim and must avoid contradictions across sections.

Create a decomposition-based workflow that the LLM (and/or tools) will follow. Your design must include:
1) An initial set of sub-problems (as explicit questions) that are solved sequentially.
2) For each sub-problem, specify what context is passed in (including how you will carry forward prior Q/A pairs) and what the expected output format is.
3) At least one point where the workflow can *recursively decompose* a sub-problem if the model cannot confidently answer (e.g., conflicting evidence, missing dates, ambiguous causality). Describe the trigger condition and the recursive sub-questions.
4) A final synthesis step that combines the accumulated Q/A pairs into the narrative while enforcing consistency and traceability.

Keep your answer concrete: provide 5–8 sub-problem questions, show an example of the context bundle for sub-problem #3 (what exactly is included), and describe the recursive branch in enough detail that an engineer could implement it.

Create a Recursive, Context-Carrying Decomposition Plan for LLM-Assisted KPI Narrative Generation

You are leading an internal enablement team building an LLM-powered assistant to produce a board-ready “AI Risk & Compliance Brief” from a 60-page draft policy, 12 incident tickets, and 3 regulator updates. The brief must be accurate, traceable to sources, and consistent across sections (e.g., the risk taxonomy used in the executive summary must match the detailed findings). You have observed two failure modes in a prototype: (1) the model generates a good outline but later sections contradict earlier claims, and (2) some sub-questions (e.g., “Which incidents map to which regulatory obligations?”) are still too complex and the model gives shallow answers.

Write an essay proposing a concrete end-to-end problem-decomposition workflow that addresses these failures. Your answer must: (a) show how you will generate sub-problems (including at least one example of a well-formed sub-problem), (b) explain how you will solve sub-problems sequentially while carrying forward context using explicit Q/A pairs (include a short illustrative snippet of what the prompt/context for a later step would look like), (c) identify at least one sub-problem that should be decomposed recursively and demonstrate what the next level of sub-problems would be, and (d) justify the tradeoffs of your design (e.g., quality/consistency vs. token cost/latency, and how your approach reduces contradictions and shallow reasoning).

Designing a Decomposition-Driven LLM Workflow for a High-Stakes Corporate Task

You are leading an internal enablement team building an LLM-powered assistant to answer employee questions about a 120-page, frequently updated Benefits & HR policy PDF. The assistant must answer multi-hop questions (e.g., “If I switch from full-time to part-time mid-year, how does that affect my HSA eligibility and the employer match, and what forms do I need?”). You have a strict constraint: each user question can trigger at most 6 LLM calls, and the total context window is limited, so you cannot paste the entire policy or all prior intermediate work into every call.

A junior engineer proposes this workflow:
1) Generate a fixed list of 10 sub-questions up front for any user query.
2) Solve them in order.
3) For each step, only pass the immediately previous sub-question and answer (not the original user question or earlier Q/A pairs).
4) If a sub-question is hard, “just ask the model to think harder,” without further decomposition.

Write an evaluation of this proposed workflow and then propose a revised workflow that would be more reliable under the constraints. Your answer must (a) explain how you would generate sub-problems so they are neither too generic nor too rigid, (b) describe when and how you would apply recursive decomposition to a sub-problem that is still too complex, and (c) specify exactly what contextual information you would carry forward between sequential steps (e.g., which Q/A pairs, what summaries, what citations) and why that choice improves correctness while respecting the context-window and 6-call limits. Conclude by describing one concrete failure mode your redesign prevents and the mechanism by which it prevents it.

Evaluating and Redesigning a Decomposition Workflow Under Context and Cost Constraints

You are the product owner for an internal LLM assistant that must answer employee questions about a new travel-and-expense (T&E) policy by reading a 40-page policy PDF and producing a final, auditable answer with citations. The team implemented a decomposition workflow: (1) generate a fixed list of sub-questions up front, (2) answer them one-by-one, and (3) synthesize a final response. In pilot, the assistant often gives a confident but wrong final answer when later sub-questions depend on definitions or exceptions discovered mid-way (e.g., a term like “client entertainment” is defined in an appendix and changes how earlier answers should be interpreted). 

Write an evaluation and redesign proposal that explains: (a) how you would change sub-problem generation to be dynamic and allow new sub-questions to be created as intermediate answers reveal missing dependencies, (b) when and how you would use recursive decomposition for a sub-problem that is still too complex to answer directly (give one concrete example sub-problem and its recursive breakdown), and (c) how you would structure sequential sub-problem solving so that each step receives the original question plus prior question–answer pairs as context, including what you would include/exclude to reduce error propagation and context bloat. Your answer should make clear the causal links and tradeoffs among decomposition, recursion, and contextual QA carry-forward, and it should end with a brief description of how the final synthesis would use the accumulated sub-answers to produce an auditable response.

Debugging a Decomposition-Based LLM Workflow Using Recursive Sub-Problems and Contextual QA Pairs

You are building an internal LLM-powered assistant for Site Reliability Engineers (SREs) to produce a root-cause analysis (RCA) draft within 30 minutes of a Sev-2 incident. The assistant receives: (1) a short incident timeline written by the on-call engineer, (2) links to 6 log excerpts (plain text), and (3) a runbook. The assistant must output an RCA draft with: suspected root cause, contributing factors, customer impact, and 3 concrete follow-up actions.

Constraints: The LLM context window is limited, so you cannot paste all logs at once. The incident timeline is sometimes wrong or incomplete. Some sub-questions (e.g., “What changed in the last deploy?”) often require further breakdown (e.g., identify deploy ID → list changed services → inspect config diffs). You also need traceability: the RCA draft must cite which intermediate findings it relied on.

Case study task: Propose a decomposition-based workflow that (a) generates an initial set of sub-problems, (b) solves them sequentially while carrying forward prior Q/A pairs as context, and (c) supports recursive decomposition when a sub-problem is too complex to answer directly. Then, justify one key design tradeoff you make between (i) generating all sub-problems up front vs generating them dynamically as new evidence appears, and (ii) including all prior Q/A pairs vs selectively summarizing them to fit the context window. Your answer should be specific enough that an engineer could implement the prompting/orchestration logic.

Designing a Decomposition Workflow for Root-Cause Analysis of a Production Incident

You are implementing an internal LLM assistant to help a legal-ops team review vendor contracts for (1) data-processing obligations, (2) security requirements, and (3) termination/liability clauses. The assistant must produce a final one-page risk summary with citations to the contract text. In a pilot, the team used a single prompt and got a confident but inconsistent summary. They then tried a fixed list of sub-questions, but the model still missed a critical cross-reference: a security addendum that changes the definition of “Confidential Information,” which in turn affects the liability cap.

Design a decomposition-based workflow that (a) generates sub-problems, (b) solves them sequentially while carrying forward context as question–answer pairs, and (c) uses recursive decomposition when a sub-problem is too complex (e.g., resolving definitions and cross-references). Your answer must specify: the initial set of sub-problems you would generate, the order you would solve them in, what each step’s stored QA pair should contain to support later steps, and one example of a sub-problem that should trigger recursive decomposition (including what the second-level sub-problems would be).

Designing a Decomposition-and-QA-Pair Workflow for Contract Review with Recursive Escalation

You are leading an internal team deploying an LLM to triage inbound customer emails for a regulated financial-services product. The system must (1) classify each email into one of five queues (Billing, Technical Issue, Account Access, Complaint, Potential Fraud), (2) extract any required follow-up questions to ask the customer, and (3) produce a short agent-facing rationale that cites only information present in the email. The current design uses a fixed list of sub-problems generated up front and then solved in order. In production, you observe two recurring failures: (a) the model asks irrelevant follow-up questions because it forgets constraints discovered earlier (e.g., it already determined the customer is locked out but later asks for billing details), and (b) some emails contain nested issues (e.g., a complaint that includes a potential fraud indicator) that the fixed decomposition misses, leading to misrouting.

As the owner, propose a revised end-to-end LLM interaction plan that addresses both failures by combining: (i) how you will generate sub-problems, (ii) how you will solve them sequentially while preserving intermediate results, and (iii) when and how you will trigger recursive decomposition for a sub-problem that turns out to be too complex or multi-faceted. Your answer must include at least one concrete example of a contextual QA pair you would carry forward and explain how it changes the next sub-problem’s prompt or decision.

Learn Before

Related