Self-Correcting RAG: Enhancing Faithfulness via MMKP Context Selection and NLI-Guided MCTS

Shijia Xu; Zhou Wu; Xiaolong Jia; Yu Wang; Kai Liu; April Xiaowen Dong

cs.CLApr 12, 2026

Self-Correcting RAG: Enhancing Faithfulness via MMKP Context Selection and NLI-Guided MCTS

Authors: Shijia Xu, Zhou Wu, Xiaolong Jia, Yu Wang, Kai Liu, April Xiaowen Dong

Organizations: Chongqing University, China · Queen Mary University of London, UK · Chongqing Key Laboratory of Big Data Intelligence and Privacy Computing, China · Fangda Partners, China

Abstract

Retrieval-augmented generation (RAG) substantially extends the knowledge boundary of large language models. However, it still faces two major challenges when handling complex reasoning tasks: low context utilization and frequent hallucinations. To address these issues, we propose Self-Correcting RAG, a unified framework that reformulates retrieval and generation as constrained optimization and path planning. On the input side, we move beyond traditional greedy retrieval and, for the first time, formalize context selection as a multi-dimensional multiple-choice knapsack problem (MMKP), thereby maximizing information density and removing redundancy under a strict token budget. On the output side, we introduce a natural language inference (NLI)-guided Monte Carlo Tree Search (MCTS) mechanism, which leverages test-time compute to dynamically explore reasoning trajectories and validate the faithfulness of generated answers. Experiments on six multi-hop question answering and fact-checking datasets demonstrate that our method significantly improves reasoning accuracy on complex queries while effectively reducing hallucinations, outperforming strong existing baselines.Our code is available at https://github.com/xjiacs/Self-Correcting-RAG .

Figures & tables

Figure 1: Comparison of document retrieval and selection workflows between the Baseline (Top-k) and the proposed Self-Correcting RAG Framework.

Figure 2: Illustration of the comparison between traditional retrieval paradigms and our proposed framework.(a) The Baseline Traditional Top-K RAG relies primarily on simple relevance scoring and embeddings to select top-ranked documents within a token limit.(b) In contrast, our proposed Self-Correcting RAG Optimizer models document selection as a combinatorial optimization problem. It integrates feature extraction and a dedicated Self-Correcting Optimization Engine (employing MMKP and NLI-guided mechanisms) to efficiently select the best draft.

	NQ	PopQA	MuSiQue	2Wiki	HotpotQA	MultiHop-RAG
Num of queries	1,000	1,000	1,000	1,000	1,000	2,556
Num of passages	9,633	8,676	11,656	6,119	9,811	609

Table 1: Dataset statistics

Table 4

Method	QA		Retriev.	Faithfulness
Method	EM	F1	Recall@5	AP $\uparrow$	CR $\downarrow$	Sup
Standard RAG	25.8	36.1	49.6	0.52	0.15	0.65
w/ MMKP Only	34.5	42.3	71.8	0.58	0.13	0.71
w/ MCTS Only	31.2	39.7	50.1	0.82	0.06	0.84
Self-Correcting	37.1	45.8	72.0	0.85	0.04	0.88

Table 4: Ablations. We report the average performance across all datasets to evaluate the individual contribution of the MMKP selector and MCTS generator.

Appendix figures & tables11 assets

Supplementary material from the paper’s appendix.

Appendix

Category	Dataset	Type	Avg. Hops	# Queries	# Passages	Metric
Simple QA	NQ	Open-domain	1	1,000	9,633	EM / F1
Simple QA	PopQA	Long-tail Entity	1	1,000	8,676	EM / F1
Multi-Hop QA	MuSiQue	Multi-hop	2.40	1,000	11,656	EM / F1
	2WikiMultiHopQA	Multi-hop	2.42	1,000	6,119	EM / F1
	HotpotQA	Multi-hop	2	1,000	9,811	EM / F1
Multi-Doc QA	MultiHop-RAG	Noise Robustness	2.70	2,556	609	EM / F1

Appendix

Table 5: Statistics of the evaluation datasets. The "Queries" and "Passages" columns denote the specific counts used in our experimental evaluation.

Parameter	Value
Token Budget ( $C_{\text{token}}$ )	1500
Redundancy Budget ( $C_{\text{red}}$ )	120
Relevance Weight ( $\alpha$ )	0.7
Diversity Weight ( $\beta$ )	0.3
Similarity Threshold ( $\tau_{\text{sim}}$ )	0.82
Redundancy Cost Scale	100.0

Appendix

Table 6: Hyperparameters for the MMKP Context Selector. Weights were tuned on the HotpotQA validation set.

Generator (LLM)
Parameter	Value
Model	Qwen2.5-7B-Instruct
Temperature	0.7
Top-p	0.9
Max New Tokens	256
MCTS Search

Appendix

Table 7: Hyperparameters for NLI-Guided MCTS. The generator uses these settings during the test-time search phase.

Figure 3: Impact of token budget on retrieval performance. This figure compares the Retrieval Recall@5 of our MMKP method against RAG+MMR and Top-K baselines across varying token budgets (

C_{token}

). Our MMKP (red line) consistently outperforms the baselines, demonstrating superior robustness.

Figure 4: Complexity Analysis. Our method shows significant scaling advantages on harder queries.

Panel of Figure 5 — Figure 5: Sensitivity of Recall@5 to the redundancy budget ( $C_{red}$ ). A budget of approx. 120 provides the optimal trade-off, allowing sufficient diversity without excluding relevant evidence.

Simulations ( $N$ )	Depth ( $D$ )	F1 Score	Faithfulness (AP)
1 (Greedy)	1	36.1	0.52
8	2	40.5	0.68
16	3	43.2	0.79
24	3	45.8	0.85
32	4	46.0	0.86
64	5	46.1	0.87

Appendix

Table 8: Effect of MCTS hyperparameters on performance.

N

denotes the number of simulations, and

D

denotes the search depth. The chosen configuration (

N=24,D=3

) is highlighted.

Panel of Figure 6 — Figure 6: The prompt template used for the NLI-Guided Generator. It enforces strict adherence to retrieved context and requires explicit reasoning steps to support the MCTS verification process.

Panel of Figure 7 — Figure 7: The prompt template used for the Symbolic Plan Generator. This prompt restricts the LLM to output a structured execution plan using predefined API primitives.

Panel of Figure 8 — Figure 8: Comprehensive Trace of Failure vs. Correction. Left: The Baseline fails due to information crowding . High-similarity redundant documents about Entity A fill the context window, cutting off Entity B. The model then hallucinates to fill the gap. Right: Our approach employs MMKP (Maximal Marginal Relevance based Knapsack Problem) to filter semantic duplicates, ensuring both entities are present. Subsequently, the MCTS (Monte Carlo Tree Search) planner explores reasoning paths. It actively penalizes the hallucinated branch (Branch 1) via NLI verification and rewards the factually consistent branch (Branch 2).

Panel of Figure 9 — Figure 9: Analysis of Distractor Filtering and Numerical Verification. Left: The Baseline fails due to Distractor Crowding . Popular documents about DeepMind’s AI achievements (AlphaGo) overwhelm the context window, displacing the necessary CEO tenure dates. Right: The MMKP Selector identifies the "AI Achievement" documents as semantically redundant and removes them. This preserves space for documents containing specific tenure dates. The MCTS Planner then rejects the heuristic bias ("Founders serve longer") via NLI verification, ensuring the final answer is derived from arithmetic comparison of the retrieved dates.

Retrieval-Augmented Generation

Retrieval-augmented generation connects language models to external knowledge so answers can use current, domain-specific evidence.

See all 761 papers

Parameters vs. Context: TRACE Fine-Tuning for Robust Retrieval-Augmented Generation
Sep 24, 2026Zhengchen Huang, Yundong Sun, Minrui Song +4RAG Robustness Retrieval-Augmented Generation
School of Information Science & Engineering, LinYi University, China · School of Computer Science and Technology, Harbin Institute of Technology, China
When Context Misleads: Intent-Guided Decoding for Robust Retrieval-Augmented Generation
Aug 17, 2026Haolin Jin, Pengyue Yang, Huaming ChenRAG Citation Faithfulness Retrieval-Augmented Generation
School of Electrical and Computer Engineering The University of Sydney Sydney, NSW, Australia
Reason and Verify: A Framework for Faithful Retrieval-Augmented Generation
Mar 10, 2026Eeham Khan, Luis Rodriguez, Marc QueudotRAG Citation Faithfulness Retrieval-Augmented Generation
Concordia University, Montréal, QC, Canada · Centre de recherche informatique de Montréal (CRIM), Montréal, QC, Canada

Sep 24, 2026J/K move · Enter open · S save

Parameters vs. Context: TRACE Fine-Tuning for Robust Retrieval-Augmented Generation

Zhengchen Huang, Yundong Sun, Minrui Song, Shuanglong Yao+3

School of Information Science & Engineering, LinYi University, China · School of Computer Science and Technology, Harbin Institute of Technology, China

RAG Robustness Retrieval-Augmented Generation Knowledge Conflicts in Language Models

Open paper AskPDF arXiv

Retrieval-Augmented Generation (RAG) mitigates knowledge obsolescence and factual hallucination in large language models by introducing external context. However, when retrieved knowledge conflicts with the model's internal parametric knowledge, the model may either blindly follow misleading context or incorrectly rely on parametric knowledge, leading to unreliable responses. To address this issue, this paper proposes TRACE (Debate-TRace and Answer-Completeness rEgularized fine-tuning), a robust fine-tuning framework for RAG under knowledge conflicts. First, we propose a fine-tuning method that leverages multi-agent debate traces to extract correct candidates, incorrect candidates, and answer-shift patterns, providing fine-grained supervision for reliable knowledge-source selection. In addition, we design an answer completeness regularization mechanism to alleviate empty, overly short, and prematurely terminated responses via answer-tail token reinforcement and premature termination suppression. The fine-tuning objective combines correct-answer supervision, incorrect-candidate suppression, answer-tail token reinforcement, and premature termination suppression, enabling the model to use reliable external context, resist misleading or irrelevant retrieved content, and fall back to parametric knowledge when retrieved evidence is unreliable. Experiments across multiple knowledge-conflict scenarios and datasets show that TRACE improves robustness against misleading retrieved knowledge and reduces incomplete answers. These results demonstrate that multi-agent debate traces and answer completeness regularization jointly enhance knowledge-source selection, conflict robustness, and answer quality in RAG models. Our code is available at https://github.com/PHD-lanyu/TRACE.

Abstract

Retrieval-Augmented Generation

Explore similar work

Parameters vs. Context: TRACE Fine-Tuning for Robust Retrieval-Augmented Generation

When Context Misleads: Intent-Guided Decoding for Robust Retrieval-Augmented Generation

Reason and Verify: A Framework for Faithful Retrieval-Augmented Generation