---
title: 'ChatNDC: Climate Policy RAG System'
url: https://www.emergentmind.com/topics/chatndc
type: topic
---

# ChatNDC: Climate Policy RAG System

Searching arXiv for the cited paper and any directly relevant ChatNDC references.
ChatNDC is a climate-policy retrieval-augmented generation system examined as a domain-specific LLM product in research on inference-time energy use and output quality. In the reported architecture, a user’s natural-language question is processed through query classification, document retrieval over verified climate documents, response generation, and hallucination checking, with final output assembled only from verified sentences carrying page-level references [2604.00053]. The system is studied comparatively against ChatNetZero and direct GPT-4o-Mini usage, with emphasis on how architectural choices such as multiple LLM calls, verification design, and output length shape both energy consumption and response characteristics [2604.00053].

## 1. System role and architectural definition

ChatNDC is presented as one of two LLM-based climate analysis chatbots used to study the energy footprint of domain-specific RAG workflows relative to generic model usage [2604.00053]. Its function is climate-domain analysis grounded in verified climate documents, and its workflow is explicitly decomposed into retrieval, generation, and hallucination-checking components [2604.00053].

The pipeline begins with **query classification (GPT-4o)**. The user’s natural-language question is first passed to a GPT-4o model that determines which sub-corpora are relevant and tags the query with thematic labels. The listed examples of sub-corpora are **UNFCCC submissions, IPCC reports, national policy assessments** [2604.00053]. This stage therefore performs both routing and semantic categorization within a climate-policy evidence base.

The second stage is **document retrieval (CPU-bound)**. Based on the classification output, the system issues a vector-search query over a precomputed embedding index of verified climate documents. Retrieved chunks carry provenance metadata, specifically document and page number [2604.00053]. This provenance design is central to later citation attachment and verification.

The third stage is **response generation (GPT-4o-Mini)**. Retrieved evidence is concatenated with a database-aware prompt template, and a GPT-4o-Mini call generates a draft answer under a constrained prompt with **temperature = 0** and **max length 400–600 words** [2604.00053]. The use of a constrained prompt indicates an attempt to standardize output behavior within a bounded answer length.

The fourth stage is **hallucination checking (GPT-4o-Mini)**. Each sentence of the draft is re-embedded and sent, along with its source citations, to a second GPT-4o-Mini invocation. This “anti-hallucination” agent confirms or rejects each claim and attaches fine-grained, auditable citations [2604.00053]. The final assembly stage then includes only sentences verified by the hallucination-checking step, and per-sentence references are linked to the original pages [2604.00053].

This design differs from the comparison systems in specific ways. ChatNetZero uses a single GPT-4o-Mini generation call and a non-LLM cosine-similarity check, whereas GPT-4o-Mini baselines have no retrieval or explicit verification calls [2604.00053]. A plausible implication is that ChatNDC embodies a more agentic RAG design, with stronger procedural grounding but higher orchestration overhead.

## 2. Retrieval, generation, and verification workflow

ChatNDC’s workflow is organized around an explicit separation of evidence selection, answer drafting, and post hoc claim validation. This is not merely an implementation detail; in the reported study, it is the basis for decomposing per-query energy use and for analyzing design trade-offs [2604.00053].

In the classification stage, GPT-4o identifies relevant sub-corpora and thematic labels. This means that retrieval is not performed over a monolithic corpus in an undifferentiated fashion; instead, the retrieval scope is narrowed using an upstream LLM routing decision [2604.00053]. Because the system targets climate-policy analysis, the sub-corpus selection mechanism functions as a domain-aware control layer over the evidence store.

Retrieval itself is vector search over a precomputed embedding index of verified climate documents. The retrieved chunks preserve provenance metadata including document and page number [2604.00053]. This metadata is carried forward into answer construction and later into per-sentence references. The emphasis on verified documents and page-level provenance indicates that grounding is implemented not only through semantic retrieval but also through auditable source linkage.

During generation, retrieved evidence is combined with a database-aware prompt template and processed by GPT-4o-Mini under a constrained decoding configuration. The prompt temperature is fixed at zero, and the target output length is capped within 400–600 words [2604.00053]. These settings suggest an operational preference for determinism and bounded verbosity.

The hallucination-checking stage is distinctive. Each sentence of the draft is re-embedded and evaluated with a second GPT-4o-Mini call together with source citations. The “anti-hallucination” agent either confirms or rejects each claim and attaches fine-grained citations [2604.00053]. Only verified sentences survive into the final output. This suggests a sentence-level filtration regime rather than whole-answer acceptance or rejection.

A plausible implication is that ChatNDC implements a two-pass LLM architecture in which the first pass optimizes for synthesis from retrieved evidence and the second pass optimizes for claim-level validation. The study, however, also shows that this additional structure has a measurable energy cost [2604.00053].

## 3. Energy-estimation methodology

The study estimates ChatNDC’s inference-time energy by decomposing total per-query energy into retrieval, generation, and verification components using a refined API-based framework from Jegham et al. (2025) [2604.00053]. The governing equation is:

$$
E_{\text{total}} = E_{\text{retrieval}} + E_{\text{inference}} + E_{\text{verification}}
$$

Here, $E_{\text{retrieval}}$ estimates CPU-core energy for semantic search in Colab, $E_{\text{inference}}$ covers the GPT-4o and GPT-4o-Mini generation calls, and $E_{\text{verification}}$ covers the extra GPT-4o-Mini hallucination check [2604.00053].

For retrieval-stage energy, the reported formula is:

$$
E_{\rm retrieval}
\;=\;
T_{\rm retrieval}\;(\mathrm{h})
\;\times\; P_{\rm CPU}\;(\mathrm{kW})
\;\times\; \mathrm{PUE}_{\rm Google}
\;=\;
T_{\rm retrieval}\;(\mathrm{h})
\;\times\; 0.0085\;\mathrm{kW}
\;\times\; 1.09
\;=\;
T_{\rm retrieval}\;(\mathrm{h})\;\times\;0.00927\;\mathrm{kW}
$$

For LLM generation and verification, the study defines $T_{\rm llm}$ as elapsed API-call time in hours and introduces a node power draw term:

$$
W = U_{\rm GPU}\times P_{\rm GPU} + U_{\rm nonGPU}\times P_{\rm nonGPU}
$$

after multiplying by PUE [2604.00053]. For GPT-4o-Mini, the energy estimate is:

$$
E_{\rm GPT4o\mbox{-}Mini}
\;=\;
T_{\rm llm}\;(\mathrm{h})
\;\times\;0.343125\;\mathrm{kW}
\;\times\;1.12
\;=\;
T_{\rm llm}\;(\mathrm{h})
\;\times\;0.3843\;\mathrm{kW}
$$

and similarly for GPT-4o (large) [2604.00053].

Within ChatNDC, the study states that **$E_{\text{inference}}$ = sum of all GPT-4o-Mini generation calls (here, the second stage)**, and **$E_{\text{verification}}$ = one GPT-4o-Mini call (third stage)**, both at **$0.3843$ kW $\times$ call duration** [2604.00053]. This decomposition enables component-level attribution rather than only end-to-end energy reporting.

The experimental protocol ran **102 climate-policy queries $\times$ 4 time windows (morning, afternoon, evening, random) from the Netherlands**, converting execution times into energy via the stated formulas [2604.00053]. Geographic variation beyond a single origin was not tested, and cross-location effects remain an open question [2604.00053].

## 4. Measured energy footprint

The reported quantitative results provide median and mean per-query energy for ChatNDC over **408 runs**. The values are presented both in kilowatt-hours and in kilojoules, using **$1 \text{ kWh} = 3\,600 \text{ kJ}$** [2604.00053].

| Component | Median (kWh) | Mean (kWh) |
|---|---:|---:|
| Retrieval | 0.00010 | 0.00011 |
| Generation (GPT-4o-Mini) | 0.00280 | 0.00295 |
| Verification (GPT-4o-Mini) | 0.00163 | 0.00168 |
| Total | 0.00453 | 0.00474 |

Expressed in kilojoules, the median values are **0.36 kJ** for retrieval, **10.08 kJ** for generation, **5.87 kJ** for verification, and **16.31 kJ** in total; the mean total is **17.06 kJ** [2604.00053]. The distribution indicates that retrieval contributes a small fraction of total energy relative to the LLM stages, while verification is itself a substantial component rather than a negligible post-processing step.

Time-of-day variation was reported as small, specifically **less than 5%** [2604.00053]. The median per-query values were **0.00445 kWh (16.0 kJ)** in the morning, **0.00451 kWh (16.2 kJ)** in the afternoon, **0.00460 kWh (16.6 kJ)** in the evening, and **0.00453 kWh (16.31 kJ)** for the random run [2604.00053]. The study does not report multi-region deployment behavior; it explicitly notes that geographic variation was not tested beyond the Netherlands [2604.00053].

These measurements support the paper’s broader conclusion that the energy consumption of domain-specific RAG systems depends strongly on system design, particularly the number and type of LLM calls added for routing and verification [2604.00053].

## 5. Comparative positioning against ChatNetZero and GPT-4o-Mini

The comparative analysis reports median energy use for ChatNDC, unconstrained GPT-4o-Mini, GPT-4o-Mini with responses capped at **$\le 200$ words**, and ChatNetZero [2604.00053].

| System | Median (kWh/query) | Relative to ChatNDC |
|---|---:|---:|
| ChatNDC | 0.00453 | 100 % |
| GPT-4o-Mini (unconstrained) | 0.00113 | 25 % |
| GPT-4o-Mini ($\le 200$ words) | 0.00050 | 11 % |
| ChatNetZero (CNZ) | 0.000408 | 9 % |

The study states that ChatNDC uses **approximately 4$\times$ more energy than unconstrained GPT-4o-Mini, approximately 9$\times$ more than GPT-4o-Mini(200), and approximately 11$\times$ more than ChatNetZero** [2604.00053]. A **one-way ANOVA test across the four conditions yields $p < 0.01$**, indicating statistically significant differences in median energy use [2604.00053].

The paper attributes part of this gap to architectural differences. ChatNDC includes extra GPT-4o-Mini calls for classification and hallucination check, which account for **approximately 36% of total energy**, with **30.9% from verification alone** [2604.00053]. By contrast, ChatNetZero places verification in a **lightweight cosine-similarity post-processing stage**, accounting for **8.4% of its total energy** [2604.00053]. This establishes a concrete design contrast between LLM-mediated verification and non-LLM verification.

A plausible implication is that ChatNDC occupies a point in the design space where stronger procedural checking and richer grounding flexibility are traded for materially higher inference-time energy cost. The study explicitly frames this as a question of design trade-offs rather than a simple ranking of systems [2604.00053].

## 6. Quality trade-offs, interpretation, and open questions

The quality analysis reports that **factual accuracy (statement-level) is broadly similar across ChatNDC, CNZ, and GPT-4o-Mini, within less than $\pm 3\%$ on a 0–1 index** [2604.00053]. At the same time, **embellishment (non-factual) is higher for ChatNDC (0.62) than for CNZ (0.15)**, and the paper states that this suggests extra length or style conventions drive energy cost without proportional factual gains [2604.00053].

The study therefore concludes that **agentic pipelines with multiple LLM calls can improve grounding flexibility but substantially increase energy use for only marginal improvements in factual performance** [2604.00053]. This conclusion is tightly linked to the system’s architecture: classification, retrieval, constrained generation, sentence-level hallucination checking, and final filtering all contribute to a more elaborate workflow whose added cost does not map linearly onto measured factual gains.

Several design recommendations follow directly from the reported findings. Energy should be treated as a **first-class metric alongside latency and accuracy** when designing domain-specific RAG products [2604.00053]. Output length matters: even with identical models, **longer or more elaborate answers drive up energy use and often increase embellishment**, so constraining answer length is described as a simple lever to reduce both energy and non-factual content [2604.00053]. The paper also points to **dynamic routing and cascades**, naming **FrugalGPT** and **RouteLLM**, as mechanisms for directing simple factual queries to smaller models or non-LLM modules while reserving heavyweight pipelines for more complex analytic tasks [2604.00053]. It further notes that **sparsely-activated MoE architectures** may decouple capacity from energy use [2604.00053].

The open questions are stated explicitly: **multi-round dialogue footprint; cross-regional carbon intensity; end-to-end session-level energy; real-time “agentic” budget allocation** [2604.00053]. For ChatNDC specifically, these unresolved issues delimit the current scope of evidence. This suggests that the reported measurements characterize a single-query inference setting rather than the full operational footprint of sustained interactive use.

In that sense, ChatNDC is significant not only as a climate-policy RAG system, but also as a concrete case study in how domain-specific LLM product design affects measurable energy use, verification behavior, and output form [2604.00053].

Source: https://www.emergentmind.com/topics/chatndc