Parameters vs. Context: TRACE Fine-Tuning for Robust Retrieval-Augmented Generation
Authors: Zhengchen Huang, Yundong Sun, Minrui Song, Shuanglong Yao, Ye Liu, Ji Chen, Xing Wang
Organizations: School of Information Science & Engineering, LinYi University, China · School of Computer Science and Technology, Harbin Institute of Technology, China
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.
Figures & tables
Fig. 1: Performance under different knowledge-conflict contexts on ConFiQA-MR, ConFiQA-SC, and ExplainPE.
Fig. 2: Overview of TRACE. Debate is used only during data construction to expose source-selection errors and answer shifts. The target model is then fine-tuned with debate-trace supervision and answer completeness regularization.
Knowledge-source label
Functional type
Construction condition
Training role
internal / external / both
Reliable supervision
Golden answers or reliable correct candidate answers observed during debate.
Supervises correct answers and preserves basic answering ability.
Correct-to-wrong prevention
A correct candidate appears in an intermediate round, but the later or final response drifts to an incorrect answer.
Suppresses shifts from correct candidates to incorrect answers and improves answer stability.
Boundary correction
The candidate is close to the correct answer but is overly short, boundary-unclear, or difficult to judge.
Improves answer completeness, parsability, and final evaluation determinability.
external
Context usage
The retrieved context supports the golden answer, but the query-only response is incorrect.
Trains the model to actively use external evidence when it is reliable.
internal
Context resistance
The query-only response is correct, but the retrieved context contains misleading or irrelevant.
Trains the model to avoid blindly following erroneous retrieved content.
both
Dual-source hard negative
Both parametric knowledge and retrieved context support the correct answer, but incorrect candidates still appear during debate.
Reduces the probability of abnormal incorrect candidates under dual-source reliable conditions.
TABLE I: Taxonomy of fine-tuning samples for TRACE.
Question
Incomplete output
Complete answer
What position does the spouse of the head of state of the country where Heath Ledger is a citizen hold?
Consor
Consort of the United Kingdom
What position is held by the spouse of the head of government of the country where John F. Kelly is a citizen?
First
First Lady of the United States
What is the historical significance of Fort Dearborn in relation to the creator of The Mandalorian ?
Fort Dear
Fort Dearborn
TABLE II: Examples of incomplete answers without answer completeness regularization.
S1: Correct
S2: Wrong
Method
PC-MR
PC-MC
PC-QA
NC-MR
NC-MC
NC-QA
Qwen2.5–7B
Query-only prompting
27.72%
24.66%
31.67%
25.93%
25.82%
32.31%
RAG prompting
65.68%
66.39%
74.35%
13.47%
8.06%
11.31%
CK-PlUG
64.69%
66.55%
78.66%
11.62%
8.06%
7.92%
Astute-RAG
65.51%
66.05%
77.62%
12.79%
7.07%
10.34%
TABLE III: EM (%) in standard conflict scenarios. The best result is in bold, and the second-best result is underlined.
S3: Conflict
S4: Irrelevant
S5: Partly Irrelevant
Method
SC
ExplainPE
HotPotQA
2Wiki
MuSiQue
Qwen2.5–7B
Query-only prompting
29.67%
64.45%
20.90%
25.54%
4.36%
RAG prompting
59.50%
62.21%
20.36%
22.53%
6.41%
CK-PlUG
55.00%
55.00%
22.74%
24.76%
6.25%
Astute-RAG
54.20%
56.74%
17.87%
20.35%
6.29%
TABLE IV: EM (%) in more complex retrieval scenarios. The best result is in bold, and the second-best result is underlined.
Variant
PC-MC
NC-MC
PC-MR
NC-MR
PC-QA
NC-QA
Full TRACE
91.05%
39.04%
90.92%
47.64%
86.91%
37.48%
w/o debate
85.14%
35.86%
86.30%
44.94%
87.26%
36.99%
w/o strong-teaching-weak
92.06%
38.15%
87.45%
45.45%
85.54%
35.86%
w/o answer completeness regularization
85.97%
38.65%
83.00%
42.59%
85.88%
32.14%
TABLE V: Ablation results on ConFiQA EM (%). The best result in each column is in bold.
Fig. 3: Role-conversion results between stronger debate models and a weaker target model, with comparison to the strongest baseline.
Retrieval-augmented generation (RAG) improves large language models (LLMs) by incorporating external evidence, but it also introduces knowledge conflicts when retrieved contextual knowledge (CK) and parametric knowledge (PK) disagree or are both unreliable. Existing approaches mainly coordinate which source to use, without explicitly asking whether each answer path is correct. We argue that faithful RAG requires LLM self-awareness, namely the ability to recognize the limits of its own knowledge and reasoning. To ground this problem, we construct a model-specific, ground-truth-aligned knowledge-conflict benchmark by evaluating LLM backbones on PK-only and CK-conditioned answer paths over approximately 69K query-context instances per backbone, drawn from five conflict-QA datasets. We then introduce SABER, a Self-Aware Belief Estimator for RAG that requires no LLM fine-tuning. SABER combines a self-prior with PK-side and CK-side conditional reasoning representations from multi-trace inference, then estimates reliability beliefs with two lightweight predictors to drive a 4-cell decision over trust PK, trust CK, trust either, or abstain. Across four LLM backbones, SABER improves end-to-end accuracy and conflict-specific faithfulness over ten inference-time and fine-tuning baselines, with the largest gains on conflict-heavy datasets. Under abstention, SABER's risk-coverage curve Pareto-dominates every prompt-based abstainer, providing a tunable balance between coverage and answer risk. Our code is available at https://github.com/xizhu1022/SABER.
Xi Zhu, Ziqi Wang, Kai Mei +5
Rutgers University · University of Illinois Urbana-Champaign
Retrieval-augmented generation (RAG) improves large language models by grounding generation in external evidence, but it also introduces a source trust problem: retrieved context may be useful, irrelevant, or even misleading. Existing RAG systems often apply a fixed trust policy toward retrieved evidence, which can either over-trust incorrect context or underuse context when the user explicitly asks for context-following behavior. Therefore, we propose Intent-Guided Decoding (IGD), a framework that arbitrates between retrieved context and parametric memory according to user intent. IGD uses answer-level filtering and token-level correction to steer the final decoding trajectory between retrieved context and parametric memory. We evaluate IGD on three faithful QA benchmarks and three factual-conflict benchmarks across five LLMs, IGD substantially improves factual recovery, achieving gains of up to 65.4 percentage points on factual-conflict benchmarks over Direct RAG, while preserving or improving strict context-following behavior, this findings highlight the importance of balancing factuality and faithfulness in RAG.
Haolin Jin, Pengyue Yang, Huaming Chen
School of Electrical and Computer Engineering The University of Sydney Sydney, NSW, Australia
Retrieval-Augmented Generation (RAG) significantly enhances Large Language Models (LLMs) by providing access to external knowledge. However, current research primarily focuses on retrieval quality, often overlooking the critical ''integration bottleneck'': even when relevant documents are retrieved, LLMs frequently fail to utilize them effectively due to conflicts with their internal parametric knowledge. In this paper, we argue that implicitly resolving this conflict in a single generation pass is suboptimal. We introduce GuarantRAG, a framework that explicitly decouples reasoning from evidence integration. First, we generate an ''Inner-Answer'' based solely on parametric knowledge to capture the model's reasoning flow. Second, to guarantee faithful evidence extraction, we generate a ''Refer-Answer'' using a novel Contrastive DPO objective. This objective treats the parametric Inner-Answer as a negative constraint and the retrieved documents as positive ground truth, forcing the model to suppress internal hallucinations in favor of external evidence during this phase. Finally, rather than naive concatenation or using the DPO trained model directly, we propose a joint decoding mechanism that dynamically fuses the logical coherence of the Inner-Answer with the factual precision of the Refer-Answer at the token level. Experiments on five QA benchmarks demonstrate that GuarantRAG improves accuracy by up to 12.1% and reduces hallucinations by 16.3% compared to standard and dynamic RAG baselines.
Zhengyi Zhao, Shubo Zhang, Zezhong Wang +7
The Chinese University of Hong Kong · Ministry of Education Key Laboratory of High Confidence Software Technologies, CUHK · University of International Relations +2