Not All Test-Time Compute Is Evidence: Model-Specific Bias in LLM Reasoning

Ming Shen; Tianjun Yao

cs.AIOct 16, 2026

Not All Test-Time Compute Is Evidence: Model-Specific Bias in LLM Reasoning

Authors: Ming Shen, Tianjun Yao

Organizations: Shanghai Jiao Tong University · Shenzhen Loop Area Institute (SLAI)

Abstract

Test-time scaling methods implicitly assume that additional computation from the same model provides independent evidence. We show that this assumption fails within a single reasoning model: errors are correlated across repeated samples, across reasoning depth, and in the model's own self-evaluation, so more compute from the same model is not more evidence. We introduce DIA (Disagreement-triggered Independent Arbitration), which treats disagreement between an intermediate-stage and the final-stage prediction as a per-question signal of reasoning instability and resolves it by asking a small panel of heterogeneous models to solve the question independently, without seeing either candidate, and aggregating their answers directly. Across seven benchmarks and two base models, DIA outperforms full chain-of-thought in 12 of 14 settings, including a 7.00% gain on MuSiQue, while activating on only $\sim 17\%$ of examples and using up to 47.2% fewer tokens than a multi-agent baseline. Ablations show that both independent solving and solver heterogeneity are necessary for these gains.

Figures & tables

Figure 1: Three manifestations of model-specific error bias within a single reasoning model. (a) When repeated sampling produces an incorrect majority answer, erroneous samples concentrate on the same wrong answer rather than spreading across alternatives. (b) When a model’s initial answer is incorrect, reasoning frequently preserves the same wrong answer rather than correcting it. (c) When answers at different reasoning stages disagree, the model’s self-evaluation systematically favors the later, more elaborate outcome, even when the earlier answer is correct. Detailed analyses are in Section 2 .

Figure 2: Generation bias and initial-answer persistence bias. Left: Wrong-answer concentration under SC@20 failures. Each point denotes a failed question and reports its dominant wrong-answer share $C_{q}$ ; dashed lines mark the uniform-choice baseline. Right: Initial-answer persistence across base-model scales. Error Retained denotes the fraction of initially incorrect answers for which full CoT returns the same incorrect answer.

Dataset	Final-stage Pref.	Self-select Acc. $\uparrow$	Intermediate $\checkmark$ / Final $\times\uparrow$	Intermediate $\times$ / Final $\checkmark\uparrow$	Wrong-choice Conf. $\downarrow$
MuSR	97.7%	44.1%	0.0%	93.8%	100.0%
MMLU-Pro	95.1%	63.4%	6.7%	96.2%	100.0%

Table 1: Self-selection on reasoning-stage disagreements with Qwen3-8B. We report preference for the final-stage prediction, overall accuracy when exactly one candidate is correct, directional accuracy conditioned on which stage is correct, and confidence on incorrect selections. Wrong-choice Conf. is the mean softmax probability assigned to the chosen answer token when the selection is incorrect.

Figure 3: Overview of Dia . The base model produces an intermediate-stage prediction $a_{E}$ and a final-stage prediction $a_{F}$ . Agreement returns the shared answer directly; disagreement activates three heterogeneous solvers that solve the original question independently behind an information firewall. Answer-level arbitration aggregates support for $a_{E}$ , $a_{F}$ , or a third answer.

Models	Methods	Knowledge-intensive				Contextual & Multi-hop						Mathematical		Narrative / Soft		Average
		MMLU-Pro		GPQA-Diamond		DetectBench		MuSiQue		2Wiki		MATH		MuSR		All Datasets
		Acc. $\uparrow$	Tok. $\downarrow$	Acc. $\uparrow$	Tok. $\downarrow$	Acc. $\uparrow$	Tok. $\downarrow$	Acc. $\uparrow$	Tok. $\downarrow$	Acc. $\uparrow$	Tok. $\downarrow$	Acc. $\uparrow$	Tok. $\downarrow$	Acc. $\uparrow$	Tok. $\downarrow$	Acc. $\uparrow$	Tok. $\downarrow$
Qwen3-4B	CoT	68.00	1.00	51.01	1.00	67.74	1.00	59.00	1.00	89.50	1.00	96.67	1.00	63.50	1.00	70.77	1.00
	DEER	63.14	0.86	56.06	1.60	61.65	0.67	51.67	0.84	88.00	0.29	95.33	1.18	58.50	0.68	67.76	0.87
	Dynasor	63.71	1.68	51.01	1.76	65.95	1.30	55.00	1.16	90.50	1.21	95.67	2.00	63.50	1.51	69.33	1.52
	MAS-Exchange	70.29	4.38	59.09	2.75	67.74	4.05	61.33	5.32	90.00	3.53	97.00	2.50	62.50	4.04	72.56	3.80

Table 2: Main results across model scales. For each dataset, adjacent columns report accuracy (%) and the total input and output token ratio relative to CoT. Average Acc. and Tok. are the corresponding unweighted arithmetic means across the seven datasets. Bold indicates the highest reported accuracy in each column within each base-model block.

Setting	Dataset	Final Pref.	Sel. Acc. $\uparrow$	Inter. $\checkmark$ / Final $\times\uparrow$	Inter. $\times$ / Final $\checkmark\uparrow$
Full Trace	MuSR	93.0%	47.1%	5.6%	93.8%
Full Trace	MMLU-Pro	91.8%	73.2%	26.7%	100.0%
Equal Budget	MuSR	69.8%	50.0%	33.3%	68.8%
Equal Budget	MMLU-Pro	61.7%	65.0%	57.1%	69.2%

Table 3: Direct trace comparison with full traces or equal-budget summaries (Qwen3-8B base model). Final Pref. is the fraction of reasoning-stage disagreements where the reviewer ensemble selects the final-stage prediction. Sel. Acc. is the correct-selection rate when exactly one candidate is correct. The last two columns report the probability of selecting the intermediate-stage prediction when only it is correct, and selecting the final-stage prediction when only it is correct, respectively.

Dataset	Hetero.	3 Qwen3-8B	$\Delta$
MuSR	55.81%	51.16%	$-4.65\%$
2WikiMultiHopQA	53.85%	19.23%	$-34.62\%$
MuSiQue	54.72%	24.53%	$-30.19\%$

Table 4: Solver heterogeneity and base-model scaling. Left: Effect of solver heterogeneity with Qwen3-8B as the base model. Arbitration accuracy is measured on activated reasoning-stage disagreements; the homogeneous panel uses three independently sampled Qwen3-8B solvers. Right: Scaling the base model with a fixed heterogeneous solver panel. We report full-dataset accuracy and activation rate, defined as the fraction of examples that trigger arbitration through reasoning-stage disagreement.

Appendix figures & tables8 assets

Supplementary material from the paper’s appendix.

Appendix

Dataset	Qwen3-4B	Qwen3-8B
DetectBench	20.79%	17.92%
MMLU-Pro	19.43%	17.43%
GPQA-Diamond	2.02%	1.01%
MATH L3	9.00%	5.00%
MATH L4	5.00%	6.00%
MATH L5	9.00%	14.00%

Appendix

Table 5: Activation rates. Each entry is the percentage of evaluation examples for which the intermediate-stage and final-stage predictions disagree and arbitration is invoked.

Dataset	Activated	3 Reviewers	5 Reviewers	$\Delta$
MuSR	43	55.81% (24/43)	55.81% (24/43)	0
2WikiMultiHopQA	26	53.85% (14/26)	53.85% (14/26)	0
DetectBench	50	48.00% (24/50)	46.00% (23/50)	$-1$

Appendix

Table 6: Effect of increasing the reviewer ensemble from three to five models. Results are evaluated only on activated examples, where the intermediate-stage and final-stage predictions disagree. Bold values emphasize arbitration accuracy; the smaller gray text reports the corresponding number of correct decisions.

\Delta

denotes the change in the number of correct decisions.

Figure 4: A MAS-Exchange failure case on DetectBench. Round 1 contains two correct independent predictions (D) and two incorrect predictions (C). After all agents exchange their responses, the two initially correct agents switch from D to C, while only Qwen3-4B switches from C to D. GPT-4o-mini remains at C, producing a three-to-one incorrect majority in Round 2. Green and red text denote correct and incorrect reasoning or predictions, respectively; the arrows summarize each agent’s answer transition.

Figure 5: The corresponding Dia decision on the same DetectBench question. The intermediate-stage prediction is C with high confidence, while the final-stage prediction is the gold answer D. Candidate-blind independent solvers vote D, D, and C, and answer-level arbitration selects the final-stage prediction by a two-to-one majority. Unlike MAS-Exchange, the solvers do not observe one another’s responses before the vote. Green and red text denote correct and incorrect reasoning or predictions, respectively.

Figure 6: A qualitative Dia example on MMLU-Pro (Question 10051; Qwen3-4B base model).

Figure 7: A qualitative Dia example on MuSR (Question 109; Qwen3-8B base model).

Figure 8: A qualitative Dia example on GPQA-Diamond (Question recmkvk6EFAmqyMxR ; Qwen3-8B base model).

Figure 9: A qualitative Dia example on MATH Level 3 (Question 200; Qwen3-4B base model).

Test-Time Scaling

Test-time scaling studies how extra inference-time computation can improve model answers, decisions, and generated outputs.

See all 223 papers

Divergent-Convergent Reasoning: Scaling Test-Time Compute through Structured Solution Synthesis
Aug 15, 2026Bo Wen, Yuhao Chen, Erhan Bilal +3LLM Reasoning Inference-Time Scaling
Enkira.ai, USA · School of Computing, Queen’s University, Kingston, Ontario, Canada · IBM T.J. Watson Research Center, USA +2
Scaling Evaluation-time Compute with Reasoning Models as Evaluators
Mar 25, 2025Seungone Kim, Ian Wu, Jinu Lee +8LLM Evaluation Language Model Generation Evaluation
CMU · UIUC2 · KAIST AI3 +3
Test-Time Scaling in Reasoning LLMs: Inference Regimes, Evaluation, and Reproducibility
Aug 4, 2026Mohsen Hariri, Weicong Chen, Nahal Shahini +11LLM Evaluation Test-Time Scaling
Case Western Reserve University

Aug 15, 2026J/K move · Enter open · S save

Divergent-Convergent Reasoning: Scaling Test-Time Compute through Structured Solution Synthesis

Bo Wen, Yuhao Chen, Erhan Bilal, Carla Agurto Rios+2

Enkira.ai, USA · School of Computing, Queen’s University, Kingston, Ontario, Canada · IBM T.J. Watson Research Center, USA+2

LLM Reasoning Inference-Time Scaling Efficient Language Model Reasoning Inference-Time Compute Allocation

Open paper AskPDF arXiv

Test-time compute can substantially improve Large Language Model (LLM) reasoning performance, yet how and when additional compute helps remains poorly understood. We study Divergent-Convergent Reasoning (DCR), a simple two-phase primitive consisting of an exploration phase that generates multiple candidate solutions followed by a convergent reconciliation phase. We present three core results. First, we show that even a single reconciliation step can reliably amplify correct minority reports: across datasets, DCR often recovers the correct answer when correct exploration outputs are in the minority, a regime where majority voting fails. Second, we introduce recursive DCR, an autoregressive reconciliation system that iteratively analyzes disagreements and allocates additional test-time compute. Recursive DCR achieves higher accuracy than fixed-compute baselines-reaching 93.3% on AIME 2024 and 92.0% on AIME 2025-while using roughly 27% less compute on average, demonstrating that attentive resource allocation is superior to uniform scaling. Third, we analyze disagreement among exploration outputs via a simple, training-free dispersion metric. Dispersion reveals a structured relationship between disagreement and test-time gains: in regimes where DCR is effective, higher disagreement among exploration outputs is associated with larger accuracy improvements from reconciliation. Together, these results show that disagreement, often viewed as noise, can be systematically exploited to improve test-time reasoning and reveal emerging scaling laws for agentic LLM systems.

Abstract

Test-Time Scaling

Explore similar work

Divergent-Convergent Reasoning: Scaling Test-Time Compute through Structured Solution Synthesis

Scaling Evaluation-time Compute with Reasoning Models as Evaluators

Test-Time Scaling in Reasoning LLMs: Inference Regimes, Evaluation, and Reproducibility