Realm platform (log in using expert.micro1.ai email through Google)
Form for submitting links to your documents for review and final submission (after submitting using Realm platform)
For each assigned task:
Open the Arden task.
Review:
Full narrative prompt and legend with 20 inputs
Golden Response
Golden Response Rubric
Golden CoT Rubric
Focus on:
The 8 technology inputs ([6] through [13]) and related narrative in the prompt.
The 5 location inputs as they might affect technical assumptions (climate or utility, for example).
Rubric items that mention the technology if needed.
Your first job is to decide: “Is this scenario physically plausible if a developer tried to build it?”
Does the PV kW make sense for the property type and roof description?
Example: 8–10 kW on a typical single-family home with decent roof area is fine.
30 kW on a tiny townhouse with severe shading is likely unrealistic.
Is the expected annual generation (kWh) consistent with:
PV system size
Location (latitude, climate)
Typical performance ranges?
If PV size vs roof, or PV size vs kWh, look clearly off, propose alternative values and include in feedback.
Check whether:
Main panel rating is realistic for the building and new loads.
Battery size (kWh and kW) is reasonable given PV size and building type.
Heat pump COP is plausible for the climate and retrofit context.
EV charger rating (kW) matches real products and is reasonable for the panel capacity.
If the scenario would clearly require panel upgrades, extreme load management, or would be non-compliant with normal electrical standards, note that explicitly with a recommendation (for example, “Panel upgrade or smaller EVSE needed”).
Confirm that:
The utility service territory matches the stated city or region.
On its face, the Energy Community status is not obviously incompatible with the location (you do not have to validate the classification in detail, just catch blatant mismatches).
If there is a glaring mismatch (for example, wrong utility for the city), call it out.
Walk through the Legend and narrative:
The 8 technology inputs should form a coherent system:
PV kW, kWh, battery kWh and kW, heat pump COP, EV charger power, Domestic Content percentage, and sizing strategy should make sense together.
You do not have to re-price projects line by line, but if something is clearly unreasonable (for example, $10,000 for a large PV + battery + heat pump + EV package), call it out.
For any major inconsistency:
Suggest specific corrections (for example, “Total cost should be more like $60,000–$80,000 in this market”).
Note in feedback if the fix is large.
You do not own the rubric, but you do need to validate technical correctness for the parts that touch your domain.
Focus on rubric rows that mention the energy system itself.
For each such row:
Check technical realism:
Does the criterion describe performance or behavior that is consistent with the system defined in the prompt?
Example: A row might require the model to state PV capacity and annual kWh roughly equal to the prompt values.
Check alignment with the inputs:
If PV is 8.5 kW and 12,750 kWh/year in the prompt, the rubric should expect those values or a plausible tolerance, not a totally different number.
If the EV charger is 9.6 kW Level 2, the rubric should not reward something clearly inconsistent (for example, 3 kW).
Propose improvements:
If a criterion is too vague, unrealistic, or misaligned with the inputs, suggest a clearer wording or corrected value.
Do not directly edit the rubric yourself; your role is to annotate and recommend.
If a technical rubric row is fundamentally wrong or would systematically reward incorrect answers, note that in your feedback and explain why.
Now check the CoT rubric rows that involve technical reasoning.
Ask:
Does the rubric encourage the model to reason properly about:
PV size and kWh vs location?
Battery use (backup, arbitrage, resilience)?
Heat pump COP and its impact on usage or savings?
EV charger load and impact on panel capacity or energy?
Budget vs system sizing tradeoffs?
If important technical aspects are present in the prompt but never appear in CoT criteria, suggest adding a row that forces the model to acknowledge and use that dimension (for example, “Uses PV kW and typical kWh per kW per year for region to check annual production”).
For each CoT row you review:
Check that the reasoning step is valid:
No pseudo-physics or impossible performance assumptions.
Check that it is tied to actual inputs:
The rubric should not encourage the model to hallucinate new system specs instead of using the ones in the prompt.
Recommend missing reasoning steps:
For example, if battery is critical in the scenario but there is no CoT criterion about storage, ask for one.
Note in feedback if major technical reasoning steps are wrong or missing in a way that would systematically mis-train the model.
If there are no technical rubric items, that is okay.
For each task:
Write a short QA note (1–3 paragraphs) covering:
Whether the scenario is technically feasible
Any changes you recommend to:
Technology values (PV size, kWh, battery, EVSE, heat pump)
General energy project scope
Any items you noted in your feedback and why
Upon completion, click Action > Mark as Rework, even if the outputs require no changes.
The tax expert will then incorporate your recommendations and resolve any escalations.
You made it to the end of the instruction list!
There's a lot to take in, but these tasks should become easier with repetition.
We encourage you again to work closely with the sample feasibility assessment, which responds to the Task Prompt Example and its associated set of rubrics. Remember -- make sure the system is technically feasible using your expertise, and only return comments when something requires clarification or a change.