---
title: 'Self-NLEs: Model-Generated Explanations'
url: https://www.emergentmind.com/topics/self-natural-language-explanations-self-nles
type: topic
---

# Self-NLEs: Model-Generated Explanations

Self-Natural Language Explanations (self-NLEs) are model-generated, free-form justifications of a model's own outputs, produced as part of or immediately after decision-making. They constitute a primary mechanism by which large language models (LLMs) and other advanced neural systems seek to provide transparency, interpretability, and user-aligned rationalization for their predictions across a wide array of domains, including NLP, program synthesis, vision-language reasoning, and structured data translation. This article systematically reviews the principles, architectural variants, faithfulness challenges, evaluation strategies, and empirical impact of self-NLEs in current literature.

## 1. Definitions, Scope, and Motivations

Self-NLEs are distinct from post-hoc explanations and human-annotated rationales: they are produced by the model itself, usually conditioned on its internal representations and decision trajectory, often during inference or as a joint product with its prediction. They may take the form of causal chain-of-thought, feature- or token-level attribution, or high-level summaries referencing model-relevant latent knowledge [2106.13876, 2505.22823, 2310.11207].

Technical motivations for self-NLEs include:

- **Transparency**: Making opaque decision processes accessible in human-like terms [2106.13876, 2310.11207].
- **Faithfulness**: Ensuring the explanation accurately reflects internal model reasoning, rather than being a plausible but ungrounded post-hoc rationalization [2505.22823, 2506.07523, 2506.09277].
- **Support for end-task performance**: Providing actionable, intermediate representations for tasks such as multilingual code translation or complex question answering [2311.07070, 2411.02948].
- **Human-AI collaboration**: Facilitating user trust and error detection in decision-intensive domains [2411.02948].

Variations exist from tightly coupled joint prediction-explanation models [2203.05081], to iterative self-refinement frameworks [2505.22823], to architectures that explicitly encode and align extractive rationales, knowledge grounding, and free-text generation [2106.13876].

## 2. Principles and Architectural Variants

Self-NLE generation frameworks divide into several main paradigms:

- **Joint Prediction and Explanation**: Architectures like NLX-GPT [2203.05081] and e-UG [2105.03761] generate predictions and explanations via a single generative process, often employing shared representations to tightly couple decision and justification.
  
- **Post-hoc Rationalization with In-Loop Feedback**: CycleSQL [2411.02948] generates self-NLEs as feedback, uses them to semantically verify candidate outputs (e.g. SQL queries), and iterates until explanation/decision alignment is achieved.

- **Latent Variable Models**: ELV and variants [2011.05268] model explanations as continuous latent variables jointly with labels, optimizing variational lower bounds via alternating inference and generative steps.

- **Self-Critique and Iterative Refinement**: SR-NLE [2505.22823] employs multi-round, model-internal critique and refinement, using either natural language or feature attribution signals to iteratively improve explanation faithfulness.

- **Knowledge-Grounded Rationalization**: The RExC framework [2106.13876] grounds explanations in both extractive rationales (salient input spans) and selected background-knowledge snippets, training all modules end-to-end for maximal alignment.

- **Self-Critical Learning in Vision-Language**: S³C [2309.02155] introduces a reward that scores explanations by their causal consistency with answers, enabling semi-supervised learning without fully annotated explanations.

Representative architectural motifs and their primary characteristics are summarized below.

| Model/Framework | Coupling                  | Explanation Form         | Key Mechanism               |
|-----------------|--------------------------|-------------------------|-----------------------------|
| NLX-GPT, e-UG   | Joint                    | Free-form NLE           | Unified encoder–decoder     |
| CycleSQL        | Iterative self-verifier   | Data-grounded NLE       | SQL rewriting + NLI check   |
| RExC            | End-to-end multi-branch   | NLE + rationale + kn.   | Latent selectors, grounding |
| SR-NLE          | Iterative post-hoc        | Faithful NLE            | Critique & feedback         |
| ELV             | Latent variable EM        | Latent NLE              | Mutual optimization         |
| S³C             | Semi-supervised, RL loop  | VQA rationales          | Self-critical learning      |

## 3. Faithfulness and Self-Consistency

A central challenge for self-NLEs is the alignment between the explanation and the model’s true reasoning process. Key findings and technicalities include:

- **Faithfulness Criteria**: An explanation is faithful if and only if it mentions only those features, evidence, or intermediate steps that the model actually uses to produce its prediction [2506.09277, 2506.07523].
  
- **Quantitative Faithfulness Comparisons**: CycleSQL utilizes a feedback verifier based on NLI entailment between self-NLE and the original NL question—only accepting candidate output if semantic entailment is established [2411.02948]. SR-NLE applies counterfactual-based metrics—measuring if the explanation updates when single input tokens change the answer [2505.22823].

- **Self-Consistency of Explanations**: PSCB [2506.07523] defined self-consistency via feature-attribution vector similarity between the decision and the explanation. Standard cosine-based metrics tend to collapse the quality range; Spearman rank correlation offers greater sensitivity.

- **NeuroFaith Framework**: Recent work directly contrasts generated self-NLEs against concepts decoded from intermediate neural activations (e.g., residual stream, MHA blocks), quantifying local and global agreement in multi-hop reasoning tasks [2506.09277].

- **Empirical Gaps**: Even large, instruction-tuned models exhibit significant deviance (NLE-f ≈ 23% vs NLE-c ≈ 50–60%), with "explainer parrots"—where the explanation is correct but unfaithful—comprised 37–40% of cases [2506.09277].

- **Learning for Faithfulness**: Direct Preference Optimization (DPO) using alignment-based preferences (best/worst explanations by Spearman correlation to internal importances) can improve explanation–reasoning consistency without harming task accuracy [2506.07523].

## 4. Evaluation Strategies and Benchmarks

Self-NLEs are subject to diverse evaluation protocols:

- **Automatic Metrics**: Comprise both general NLG scores (BLEU, ROUGE, CIDEr, BERTScore) and explanation-specific measures, such as:
    - Feature/comprehensiveness: accuracy drops from masking out most-important tokens [2106.13876, 2310.11207].
    - Sufficiency: confidence gap when only most-important tokens are kept [2310.11207].
    - Consistency: attribution overlap (cosine, Spearman) between decision and explanation [2506.07523].
    - Counterfactual Tests: faithfulness under local, behavioral input perturbations [2505.22823].

- **Human Evaluation**: Quality, trustworthiness, and interpretability are rated in both task-specific settings (e.g., explaining VQA answers in e-ViL [2105.03761], SQL query outputs in CycleSQL [2411.02948]) and generic scoring (coherence, informativeness).

- **Self-Evaluation Frameworks**: Some approaches (e.g., “explain-predict” in NLX-GPT [2203.05081]) train auxiliary classifiers to check whether the explanation alone allows prediction recovery, while "retrieval-based attacks" probe for data/model bias in explanation generalization.

- **Benchmarks/Datasets**: e-ViL [2105.03761] and e-SNLI-VE [2105.03761] for vision-language, PSCB [2506.07523] for post-hoc LLM QA explanations, MultiPL-C2C [2311.07070] for code translation, eSNLI [2106.13876, 2505.22823, 2601.00282] for NLI NLEs, etc.

## 5. Domain-Specific Instantiations and Use Cases

Self-NLEs have been adapted to a range of modalities and tasks:

- **Vision-Language Reasoning**: Unified models (e.g., NLX-GPT, e-UG) couple visual reasoning, answer prediction, and explanation generation, achieving both state-of-the-art performance and high human-judged explanation quality [2203.05081, 2105.03761].

- **Program Synthesis and Translation**: "Explain-then-Translate" introduces a model-generated explanation as an intermediate representation for code-to-code translation, yielding up to +12% pass@1 gains, especially for difficult or low-resource language pairs [2311.07070].

- **Structured Data Querying**: CycleSQL introduces data-grounded self-NLEs for NL-to-SQL translation, utilizing provenance tracking and semantic rewriting to validate and improve output selection via an entailment feedback loop [2411.02948].

- **Feature Attribution in Sentiment Analysis**: ChatGPT and similar instruction-tuned LLMs can output token-level importances as self-NLEs, with faithfulness parity compared to occlusion or LIME but differing substantially in agreement structure and granularity [2310.11207].

- **Vision QA and Explanatory Multi-modal Tasks**: S³C [2309.02155] combines self-critical reward and semi-supervised learning to align free-text rationales with answer predictors, outperforming specialized models under both human and automatic metrics.

## 6. Robustness, Limitations, and Practical Guidance

Empirical investigations uncover both the robustness and shortcomings of current self-NLE frameworks:

- **Model Compression Effects**: Quantization slightly but consistently reduces self-NLE plausibility and faithfulness (up to 4.4% and 2.38%, respectively). Counterfactual explanations tend to be more robust than free-form NLEs. No quantization technique universally dominates [2601.00282].

- **Sample Efficiency and Training Regimes**: Models such as ELV [2011.05268] and SR-NLE [2505.22823] demonstrate that high faithfulness can be achieved with relatively few human explanations and, in some cases, purely self-generated feedback.

- **Trade-offs**: Performance–explanation synergy is not always realized—certain settings show explanation generation slightly degrades base task accuracy (e.g., -4% in sentiment classification [2310.11207]), and faithfulness metrics may not correlate with correctness [2506.07523].

- **Evaluation Challenges**: Existing metrics disagree, and evaluation frameworks relying on input perturbation or attribution are sometimes insensitive to explanation granularity or ill-adapted to LLMs’ behaviors [2310.11207, 2506.09277].

**Practical Recommendations** (synthesized from multiple studies [2601.00282, 2505.22823, 2411.02948]):

- Empirically validate explanation quality and faithfulness for each application and model variant.
- Consider integrating counterfactual or feature attribution–based self-critique in iterative refinement.
- Prefer explanation forms most robust to perturbations (e.g., counterfactuals for compressed models).
- Combine automatic, attributional, and human evaluations for deployment in safety-critical domains.

## 7. Conclusions and Future Directions

Recent advances in self-NLEs have produced frameworks that jointly optimize for performance, faithfulness, and user-aligned interpretability across natural language, vision-language, structured data, and programming tasks. The leading edge is moving towards tighter coupling of explanations and internal model evidence, via direct neural interpretation [2506.09277], large-scale self-consistency training [2506.07523], and data-provenance-grounded justification [2411.02948].

Future challenges include:

- **Faithfulness–Performance Coupling**: Closing the still significant gap between explanation correctness and true internal faithfulness, especially in large foundation models [2506.09277].
- **Multi-modal and Causal Faithfulness**: Extending neuro-symbolic, chain-of-thought, and concept-level metrics to multi-modal or reasoning-intensive settings [2203.05081, 2309.02155].
- **Online/Continuous Faithfulness Training**: Incorporating dynamic consistency and faithfulness signals into streaming or in-the-loop model updates [2506.07523].
- **Human-AI Coordination**: Addressing explanation utility and trust calibration in human-in-the-loop settings, especially where imperfect explanations may mislead users [2411.02948, 2310.11207].

The self-NLE paradigm is central to the interpretability agenda for LLMs and complex neural models, with ongoing research targeting greater reliability, measurable faithfulness, and seamless integration with end-task utility.

Source: https://www.emergentmind.com/topics/self-natural-language-explanations-self-nles