- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper studies how upstream messages affect downstream LLM’s answers, particularly when an incorrect message causes it to abandon an answer it could produce correctly on its own. The paper conducts controlled experiments across several benchmarks and shows the errors often follow the upstream agent’s specific wrong answer, even when the receiver has relevant evidence.
Critical strengths
-
The paper uses designed controls to distinguish the influence of message content from alternative explanations, such as source labels and additional reasoning attempts.
-
Experiments across five models and five benchmarks show that the observed failure is not limited to a single model or task.
Critical weaknesses
-
The main concern is the LLM conformity, including abandoning correct judgments under incorrect peer influence has been investigated by previous work[1]. It would be better if the paper illustrate what new conclusion this design enables beyond the prior conformity studies.
-
It is unclear how sensitive the findings are to the combination of upstream and downstream models, as the paper evaluates only selected combinations. Also, can the findings from controlled two-agent be extended to multi-agent settings?
-
The paper focuses on demonstrating substitution, but do not have much investigation into the factors that could cause it. For example, does substitution depend on the stability or uncertainty of the receiver’s original answer? Does it occur more often when evaluating the evidence requires difficult inference than when the evidence directly contradicts the upstream message? Understanding these factors would provide more insights for future work.
[1] DO AS WE DO, NOT AS YOU THINK: THE CONFORMITY OF LARGE LANGUAGE MODELS (ICLR 2025)
Questions for authors
-
Can you illustrate the novelty of the findings and the differences between this work and previous work?
-
Would different upstream and downstream model combinations lead to different conclusions? Also, can this controlled two-agent conclusion been extended to multi-agent settings?
-
Does substitution depend on the stability or uncertainty of the receiver’s original answer? Does it occur more often when evaluating the evidence requires difficult inference than when the evidence directly contradicts the upstream message?
8 comments
Response to Reviewer kVFT (2/5)
2. How sensitive are the findings to model combinations, and do they extend to multi-agent settings?
Weakness 2
It is unclear how sensitive the findings are to the combination of upstream and downstream models, as the paper evaluates only selected combinations. Also, can the findings from controlled two-agent be extended to multi-agent settings?
Question 2
Would different upstream and downstream model combinations lead to different conclusions? Also, can this controlled two-agent conclusion been extended to multi-agent settings?
2.1. Do different upstream and downstream model combinations lead to different conclusions?
Thank you for asking about sensitivity to upstream and downstream model combinations. Substitution is observed in all 20 tested combinations on each of the five tasks.
We cross four upstream models with five receivers. Substitution means a correct hidden answer changes to the upstream model's specific wrong answer when the message is shown. Adoption increase is the shown-minus-hidden difference in upstream-error adoption, in percentage points (pp).
| Measure | BIRD | HotpotQA | LBMuSiQue | 2WikiMultihopQA | DROP |
|---|---|---|---|---|---|
| Questions | 1500 | 200 | 159 | 1200 | 120 |
| Pairings with substitution | 20/20 | 20/20 | 20/20 | 20/20 | 20/20 |
| Adoption increase, pp [95% interval] | +39.01 [30.16, 49.05] | +14.73 [10.51, 19.40] | +22.13 [18.22, 25.95] | +36.61 [34.07, 39.10] | +47.16 [38.77, 55.31] |
Each cell below reports substitutions / hidden-correct answers on questions with an incorrect upstream answer (percentage).
| Upstream model | Receiver | BIRD | HotpotQA | LBMuSiQue | 2WikiMultihopQA | DROP |
|---|---|---|---|---|---|---|
| GPT-4o-mini | GPT-4o-mini | 53/99 (53.5%) | 3/62 (4.8%) | 4/13 (30.8%) | 7/38 (18.4%) | 2/2 (100.0%) |
| GPT-4o-mini | DeepSeek-V3.2 | 87/219 (39.7%) | 6/61 (9.8%) | 9/17 (52.9%) | 9/55 (16.4%) | 3/4 (75.0%) |
| GPT-4o-mini | Kimi-K2.6 | 52/220 (23.6%) | 3/63 (4.8%) | 12/19 (63.2%) | 7/45 (15.6%) | 7/12 (58.3%) |
| GPT-4o-mini | GLM-5 | 46/260 (17.7%) | 1/69 (1.4%) | 2/19 (10.5%) | 1/92 (1.1%) | 2/5 (40.0%) |
| GPT-4o-mini | Qwen3.6-Plus | 106/248 (42.7%) | 2/66 (3.0%) | 2/16 (12.5%) | 2/101 (2.0%) | 3/5 (60.0%) |
| GPT-5.4 | GPT-4o-mini | 24/89 (27.0%) | 5/5 (100.0%) | 4/9 (44.4%) | 11/11 (100.0%) | 3/3 (100.0%) |
| GPT-5.4 | DeepSeek-V3.2 | 68/117 (58.1%) | 5/5 (100.0%) | 8/8 (100.0%) | 96/107 (89.7%) | 5/6 (83.3%) |
| GPT-5.4 | Kimi-K2.6 | 75/129 (58.1%) | 3/3 (100.0%) | 6/9 (66.7%) | 41/41 (100.0%) | 11/12 (91.7%) |
| GPT-5.4 | GLM-5 | 109/158 (69.0%) | 5/6 (83.3%) | 5/10 (50.0%) | 33/155 (21.3%) | 4/6 (66.7%) |
| GPT-5.4 | Qwen3.6-Plus | 76/149 (51.0%) | 6/6 (100.0%) | 6/9 (66.7%) | 99/130 (76.2%) | 3/5 (60.0%) |
| Kimi-K2.6 | GPT-4o-mini | 50/50 (100.0%) | 7/17 (41.2%) | 1/26 (3.8%) | 9/32 (28.1%) | 3/4 (75.0%) |
| Kimi-K2.6 | DeepSeek-V3.2 | 27/84 (32.1%) | 4/14 (28.6%) | 2/29 (6.9%) | 20/87 (23.0%) | 3/6 (50.0%) |
| Kimi-K2.6 | Kimi-K2.6 | 2/47 (4.3%) | 5/10 (50.0%) | 2/21 (9.5%) | 21/71 (29.6%) | 5/8 (62.5%) |
| Kimi-K2.6 | GLM-5 | 29/53 (54.7%) | 5/18 (27.8%) | 2/28 (7.1%) | 7/125 (5.6%) | 1/8 (12.5%) |
| Kimi-K2.6 | Qwen3.6-Plus | 24/68 (35.3%) | 6/16 (37.5%) | 1/25 (4.0%) | 5/94 (5.3%) | 1/8 (12.5%) |
| Qwen-Plus | GPT-4o-mini | 13/25 (52.0%) | 6/21 (28.6%) | 2/31 (6.5%) | 45/45 (100.0%) | 1/1 (100.0%) |
| Qwen-Plus | DeepSeek-V3.2 | 22/42 (52.4%) | 3/20 (15.0%) | 2/31 (6.5%) | 38/38 (100.0%) | 3/3 (100.0%) |
| Qwen-Plus | Kimi-K2.6 | 14/36 (38.9%) | 6/21 (28.6%) | 2/30 (6.7%) | 28/38 (73.7%) | 4/7 (57.1%) |
| Qwen-Plus | GLM-5 | 23/61 (37.7%) | 4/23 (17.4%) | 4/37 (10.8%) | 38/50 (76.0%) | 2/4 (50.0%) |
| Qwen-Plus | Qwen3.6-Plus | 30/65 (46.2%) | 3/22 (13.6%) | 1/32 (3.1%) | 20/34 (58.8%) | 4/4 (100.0%) |
The expanded matrix supports the same qualitative finding beyond the combinations selected in the original submission: substitution occurs across all tested pairings, while its frequency varies across combinations and tasks.
Response to Reviewer kVFT (3/5)
2.2. Does the controlled two-agent finding extend to multi-agent settings?
We appreciate your suggestion to examine multi-agent settings. Substitution persists in every tested multi-agent configuration across the five tasks. Within each task, workflows use the same questions and incorrect drafts, with five calls per question in each shown or hidden condition.
Homogeneous workflows use GPT-4o-mini; heterogeneous workflows use GPT-4o-mini, Kimi-K2.6, Qwen3.6-Plus, and GLM-5. “Roles swapped” exchanges Kimi-K2.6 and GLM-5 between review and adjudication. Substitution is reported over all questions (n/N).
| Configuration | Workflow | Measure | BIRD | HotpotQA | LBMuSiQue | 2WikiMultihopQA | DROP |
|---|---|---|---|---|---|---|---|
| All | Questions | 88 | 354 | 110 | 120 | 120 | |
| Homogeneous | Sequential review | Substitution, n/N (%) | 60/88 (68.2%) | 28/354 (7.9%) | 71/110 (64.5%) | 20/120 (16.7%) | 20/120 (16.7%) |
| Adoption increase, pp [95% interval] | +70.45 [57.89, 83.75] | +7.91 [5.08, 10.73] | +64.55 [55.45, 73.64] | +16.67 [10.00, 23.33] | +16.67 [10.00, 23.33] | ||
| Final correct, hidden → shown | 81/88 → 20/88 | 346/354 → 232/354 | 110/110 → 27/110 | 120/120 → 100/120 | 120/120 → 100/120 | ||
| Mutual review and adjudication | Substitution, n/N (%) | 45/88 (51.1%) | 18/354 (5.1%) | 41/110 (37.3%) | 21/120 (17.5%) | 13/120 (10.8%) | |
| Adoption increase, pp [95% interval] | +51.14 [40.23, 62.50] | +5.08 [2.82, 7.34] | +37.27 [28.18, 46.36] | +17.50 [10.83, 24.17] | +10.83 [5.83, 16.67] | ||
| Final correct, hidden → shown | 82/88 → 36/88 | 354/354 → 288/354 | 110/110 → 58/110 | 120/120 → 98/120 | 120/120 → 106/120 | ||
| Independent aggregation | Substitution, n/N (%) | 33/88 (37.5%) | 11/354 (3.1%) | 40/110 (36.4%) | 14/120 (11.7%) | 12/120 (10.0%) | |
| Adoption increase, pp [95% interval] | +38.64 [28.17, 49.35] | +3.11 [1.41, 5.08] | +36.36 [27.27, 45.45] | +11.67 [6.67, 17.50] | +10.00 [5.00, 15.83] | ||
| Final correct, hidden → shown | 82/88 → 47/88 | 354/354 → 320/354 | 110/110 → 61/110 | 120/120 → 106/120 | 120/120 → 107/120 | ||
| Heterogeneous | Sequential review | Substitution, n/N (%) | 12/88 (13.6%) | 4/354 (1.1%) | 3/110 (2.7%) | 1/120 (0.8%) | 3/120 (2.5%) |
| Adoption increase, pp [95% interval] | +13.64 [6.25, 19.44] | +1.13 [0.28, 2.26] | +2.73 [0.00, 6.36] | +0.83 [0.00, 2.50] | +2.50 [0.00, 5.83] | ||
| Final correct, hidden → shown | 77/88 → 66/88 | 354/354 → 307/354 | 83/110 → 80/110 | 120/120 → 117/120 | 120/120 → 116/120 | ||
| Mutual review and adjudication | Substitution, n/N (%) | 7/88 (8.0%) | 6/354 (1.7%) | 14/110 (12.7%) | 3/120 (2.5%) | 4/120 (3.3%) | |
| Adoption increase, pp [95% interval] | +7.95 [3.30, 13.51] | +1.69 [0.56, 3.11] | +12.73 [7.27, 19.09] | +2.50 [0.00, 5.83] | +3.33 [0.83, 6.67] | ||
| Final correct, hidden → shown | 67/88 → 62/88 | 349/354 → 321/354 | 89/110 → 67/110 | 120/120 → 116/120 | 120/120 → 115/120 | ||
| Roles swapped | Substitution, n/N (%) | 14/88 (15.9%) | 8/354 (2.3%) | 6/110 (5.5%) | 6/120 (5.0%) | 5/120 (4.2%) | |
| Adoption increase, pp [95% interval] | +14.77 [4.23, 23.46] | +2.26 [0.85, 3.95] | +5.45 [1.82, 10.00] | +5.00 [1.67, 9.17] | +4.17 [0.83, 8.33] | ||
| Final correct, hidden → shown | 69/88 → 55/88 | 348/354 → 324/354 | 110/110 → 85/110 | 120/120 → 110/120 | 120/120 → 115/120 |
Every tested configuration on every task shows fewer final correct answers in the shown condition, alongside observed substitutions to the upstream model's specific wrong answer. Subsequent review, adjudication, and aggregation do not eliminate substitution in these tested settings.
Response to Reviewer kVFT (1/5)
Dear Reviewer kVFT,
Thank you for your careful review and for recognizing the value of our controlled experiments and evaluation across models and benchmarks. Your suggestions have guided us in clarifying the contribution, extending the model and workflow comparisons, and examining factors associated with substitution.
1. What new conclusions does this design enable beyond prior conformity studies?
Thank you for encouraging us to clarify the findings beyond prior conformity studies. The original submission quantifies the risk that a receiver abandons an independently correct answer for the upstream model's specific wrong answer while its independent evidence remains available. Communication gains can coexist with these directed losses on independently solvable items, so overall accuracy alone does not reveal this handoff risk.
We examine this risk through item-level shown–hidden comparisons that hold the receiver's independent evidence fixed. Task-scored correct-to-wrong (c→w) changes occur in 11.1% of independently solvable model–question cases, with a maximum of 32% in one model–benchmark combination. A separate audit of c→w transitions finds that 94% match the upstream model's specific wrong answer. Reversing the upstream conclusion while keeping the evidence fixed also changes receiver answers, providing intervention evidence that receiver answers track the message's specific conclusion.
Weng et al. [1] examine interaction protocols, interaction time, and majority size, and evaluate persona and reflection strategies for mitigating conformity. Qu et al. [2] manipulate consensus structure and authority labels, compare harmful and beneficial revisions, and test chain-of-thought and reflect-then-revise interventions.
Our supplementary experiments further quantify how directed substitution risk varies across upstream–receiver pairings and complete multi-agent workflows while independent evidence remains available. Substitution occurs across all tested model pairings, with frequencies varying across pairings and tasks (Response 2.1), and specific upstream errors can reach final answers despite subsequent review, adjudication, or aggregation (Response 2.2). Lower original-answer stability and lower self-reported confidence are associated with higher substitution rates (Response 3.1); natural-question comparisons also link the inference needed to refute the actual upstream draft with higher substitution rates (Response 3.2).