- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper proposes DENSE, a method for extracting reusable feedback from agent trajectories without access to post-execution rewards or hidden verifier outputs. DENSE constructs a nested subtask hierarchy, compresses repeated attempts into shortcuts, reconciles outstanding issues using subsequent observations, and renders feedback that emphasizes unfinished requirements while retaining selected executable details.
The paper also introduces REFIT, an evaluation protocol in which feedback methods share initial trajectories and guide separate attempts in reset environments with fresh model contexts. Experiments on Terminal-Bench 2.1 show the highest strict pass rates among the tested non-privileged feedback methods across four recipient models. GPT ablations examine hierarchy and shortcut/reconciliation components, and an exploratory extension evaluates repeated feedback on hard tasks.
Critical strengths
A useful approach to extracting partial progress without external outcome labels. DENSE distinguishes successful local discoveries from overall task completion and connects reusable evidence to outstanding requirements. This addresses a practical failure mode in long-horizon agents: useful discoveries can be lost among unsuccessful exploration or fail to translate into the required deliverable.
A reasonably controlled evaluation with consistent cross-model results. Sharing initial trajectories and resetting both environments and contexts makes comparisons between feedback methods interpretable. DENSE improves strict pass rates across all four recipients. The detailed prompts, algorithms, and outcome-transition analysis strengthen the empirical presentation, particularly by disclosing regressions and accounting limitations.
Critical weaknesses
The experiments do not sufficiently isolate the contribution of the proposed structure from actionable summarization and additional analysis computation. All-at-once performs action-level scoring, while Step-by-step evaluates decision-time justification without observing the target action’s outcome. Neither directly matches DENSE’s objective of synthesizing reusable paths and remaining requirements. The single-root summary ablation partially addresses this concern, but is evaluated only on GPT and differs from full DENSE in analysis computation and evidence organization. Consequently, the results establish the effectiveness of the complete pipeline more clearly than the necessity of its hierarchy and reconciliation mechanisms.
This concern is particularly consequential for MiniMax. Appendix D.4 reports that 32 of VF’s 37 regressions and 27 of Self-reflection’s 36 regressions involve no tool calls, whereas DENSE has no zero-tool executions. VF also repairs more initially unsuccessful executions than DENSE. Although the shared fresh-environment wrapper is a useful control, these results leave open how much of DENSE’s advantage comes from preventing premature termination rather than improving substantive recovery.
The evidence supporting the magnitude and broader interpretation of the gains remains incomplete. The study lacks an independent no-feedback retry control, limiting attribution of improvements over run0 to feedback rather than repeated attempts. This does not invalidate comparisons between feedback methods, and the appendix appropriately acknowledges the limitation for cumulative success. However, confidence intervals are also absent, despite margins of approximately 2.6–3.0 percentage points over the strongest non-privileged baselines on GPT and Kimi.
Efficiency claims require similarly careful qualification: the reported token reductions concern recipient execution only, while Figure 12 shows higher combined generation-plus-execution token use than run0 for all four recipients. The GPT dollar-cost analysis is informative, but does not establish a general reduction in total computation. Together, these limitations make the evidence insufficient for a strong claim about broadly efficient continual self-improvement.
Questions for authors
Can you provide a direct full-trajectory summarization baseline using the same generator, with matched feedback length and analysis budget, that explicitly extracts useful paths, failed approaches, supporting evidence, remaining requirements, and a rerun plan? The existing single-root ablation is relevant; extending this comparison across recipients would materially affect my assessment.
4 comments
Response to W2 (MiniMax-Specific Premature Termination vs. Substantive Recovery Gains)
- A Model-Specific Phenomenon Unique to MiniMax-M2.7:
The premature-termination factor you identified does exist on MiniMax-M2.7, and our cross-recipient audit confirms it is a phenomenon unique to MiniMax-M2.7 . Specifically, on MiniMax-M2.7, all four baselines that present linear step-by-step action–observation tables (
Self-reflection,VF,Step-by-step, andAll-at-once) incur23–37zero-tool runs—<think>traces show MiniMax mistaking linear historical steps or"all tests passed"verifier logs for actions already completed in the current session—whereas DENSE incurs 0 zero-tool runs. This also clarifies the comparison with privileged VF: having direct access to ground-truth test outputs,VFrepairs more initial failures overall (38 vs. 30), while DENSE's net lead (140 vs. 120) stems primarily from higher solution retention (110/119 vs. 82/119) and avoidingVF's 32 zero-tool regressions.
Notably, the structured Flat baseline designed in Q1 above likewise exhibits this exact non-linear-table behavior (0 zero-tool runs on MiniMax-M2.7), yet DENSE still maintains a clear +3.00 pp lead (52.43% vs. 49.44%).
-
DENSE Still Outperforms All Non-Privileged Baselines After Excluding Zero-Tool Runs: Furthermore, auditing all six main feedback methods (: 119 initial passes, 148 initial failures) and evaluating Both-Active () paired subsets—strictly excluding every run where either method makes zero tool calls—confirms that DENSE consistently outperforms every non-privileged baseline in both active pass rate and substantive failure repairs (and even among the 146 active failed-
run0pairs against privilegedVF, DENSE repairs 13 tasks whereVFactively invokes tools yet fails):Comparator on MiniMax-M2.7 Full-Sample Pass ( /267)Total 0-Tool Runs ( /267)Both-Active Pairs () Active Pass (DENSE vs. Comp) Advisor (no trace) 110 (41.20%) 0 267 / 267 140 vs. 110 (+11.24 pp) Self-reflection 103 (38.58%) 37 (27 reg.) 230 / 267 113 vs. 102 (+4.78 pp) Step-by-step 114 (42.70%) 30 (19 reg.) 237 / 267 120 vs. 114 (+2.53 pp) All-at-once 119 (44.57%) 23 (14 reg.) 244 / 267 123 vs. 118 (+2.05 pp) VF$^\dagger$ (privileged) 120 (44.94%) 34 (32 reg.) 233 / 267 108 vs. 120 (-5.15 pp) Table: MiniMax-M2.7 zero-tool counts and Both-Active () paired comparisons against DENSE (full-sample DENSE: 140/267 = 52.43%, 0 zero-tool runs, 110/119 retained, 30/148 repaired).
In the revised Appendix, we will reorganize Appendix D.4 into two separate subsections to clearly distinguish and discuss (i) this MiniMax-M2.7-specific premature-termination behavior (and its boundary relative to privileged
VF) and (ii) DENSE's consistent substantive recovery gains after excluding these zero-tool exceptions.
Response to W3 (Independent No-Feedback Retry Control and Paired 95% Confidence Intervals):
1. Independent No-Feedback Retry Control on GPT-5.5 :
Because our main evaluation (Table 2) measures single-run pass rates across all tasks in reset environments averaged over 3 repetitions (3-rep)—rather than selectively retrying failed runs or taking cumulative maxima—an unguided retry has zero expected gain over run0. To verify this directly, we ran an Independent No-feedback Retry control on GPT-5.5 across all 89 tasks 3 repetitions () under identical reset environments, fresh contexts, tools, and budgets:
| Recipient ( runs) | Baseline (run0) | No-feedback Retry | DENSE (ours) |
|---|---|---|---|
| GPT-5.5 Strict Pass (%) | 68.54% (183/267) | 68.16% (182/267, -0.37 pp) | 75.66% (202/267, +7.12 pp) |
No-feedback Retry matches run0 (-0.37 pp), whereas DENSE improves over run0 by +7.12 pp and over No-feedback Retry by +7.49 pp, confirming that the single-round gain comes from trajectory feedback rather than repeated attempts.
2. Complete Paired 95% Confidence Intervals:
From the exact paired main-experiment records (7,308 runs), we computed task-cluster paired 95% bootstrap confidence intervals ( resamples over tasks, preserving all 3 matched repetitions and shared run0 pairing within each task) for (pp):
| Comparator (DENSE Comparator) | MiniMax-M2.7 | DeepSeek V4 Pro | GPT-5.5 | Kimi K2.6 |
|---|---|---|---|---|
Baseline (run0) | +7.87 [+1.87, +14.23] | +21.81 [+13.99, +29.63] | +7.12 [+2.25, +12.36] | +11.61 [+7.87, +15.73] |
| Advisor | +11.24 [+4.87, +17.60] | +19.34 [+11.11, +27.57] | +6.37 [+0.75, +11.99] | +7.12 [0.00, +14.23] |
| Self-reflection | +13.86 [+7.87, +19.85] | +9.47 [+2.88, +16.05] | +3.37 [-1.50, +8.61] | +4.87 [+0.37, +9.36] |
| All-at-once | +7.87 [+3.00, +13.11] | +11.11 [+4.53, +17.70] | +2.62 [-2.25, +7.49] | +3.00 [-1.87, +8.24] |
| Step-by-step | +9.74 [+3.37, +16.10] | +7.82 [+1.65, +14.40] | +2.62 [-1.50, +7.12] | +4.12 [0.00, +8.24] |
| Verifier Feedback | +7.49 [0.00, +14.61] | +6.58 [-0.82, +13.99] | -8.24 [-15.73, -1.12] | -3.00 [-9.36, +3.37] |
Moreover, DENSE's gains manifest jointly across both task performance and retry execution cost/efficiency (Table 3 & Figure 4). Taking the exact strongest non-privileged baselines on GPT-5.5 and Kimi K2.6 mentioned in your comment (+2.62 to +3.00 pp pass-rate margins): on GPT-5.5, DENSE (179.9K tokens) simultaneously cuts recipient execution cost by 14.9% relative to All-at-once (211.4K) and 39.0% relative to Step-by-step (294.8K) (-43.6% vs. run0 319.1K); on Kimi K2.6, DENSE (416.1K) cuts recipient execution cost by 23.9% relative to All-at-once (546.9K) and 29.2% relative to Step-by-step (587.8K) (-23.7% vs. run0 545.3K). We will include both the No-feedback Retry control and the paired 95% CI table in Appendix.
Response to W4 (Qualifying Efficiency Claims and Cost Accounting Metrics)
Thank you for your careful reading and for pointing this out. Separating recipient retry execution cost from end-to-end pipeline computation helps place our efficiency results in clear context: reductions in recipient execution tokens (Table 3) and USD savings under a specific generator–recipient price ratio (Table 5) characterize retry-stage efficiency and analyzer–executor pricing tradeoffs rather than a universal reduction in total computation.
Specifically, our four reported metrics capture complementary aspects of cost:
- Recipient execution tokens (
run1vs.run0; Table 3 & Figure 4): Reduced by 19.0%–43.6%, counting only the recipient's retry execution (excluding feedback generation). - Combined generation + execution tokens (
Gen + run1vs.run0; Figure 12): Sums Gemini distillation tokens and recipientrun1tokens; this total is higher thanrun0across all four recipients. - Single-retry estimated USD (
Gen + run1vs.run0; Appendix B.1, Table 5): Under standard uncached base rates, Gemini feedback ($0.1396, $0.14) + GPT-5.5run1($1.0075, $1.01) totals $1.1470 ($1.15) vs. $1.8085 ($1.81) forrun0(-36.6%), reflecting a price tradeoff between a cheaper analyzer and a pricier executor rather than a universal saving across all model pairs. - Complete workflow USD (
run0 + Gen + run1; Appendix B.1): Totals $2.96 on GPT-5.5; thus, -36.6% compares a guided retry againstrun0, not the full two-round workflow from scratch against a single attempt.
In the revised manuscript, we will keep efficiency descriptions explicitly scoped to recipient execution tokens on retries with clear cross-references to the full accounting breakdown (Figure 12; Appendices A.3–A.4 & B.1), and ensure our discussion of iterative refinement remains aligned with this scope.
Thank you again for your constructive guidance—we hope the new cross-recipient Flat results (Q1) and supporting analyses fully resolve your concerns and favorably inform your assessment, and we welcome any further discussion.
Response to W1 & Q1 (Tight Coupling Between the Subtask Tree and Shortcut/Reconciliation,and Cross-Recipient `Flat` Summarization Baseline)
We sincerely thank you for your thoughtful review and constructive feedback, and for recognizing DENSE's ability to extract partial progress without external outcome labels, our controlled cross-model evaluation, and our transparent disclosure of regressions and accounting limitations.
Tight Coupling & Why Removing Hierarchy Degenerates to a Flat Call (Section 2.3–2.4, Section 4.5 & Appendix F): We have already conducted controlled ablations in Section 4.5 (Table 4) and Appendix A.2 (
w/o S&Randw/o hierarchy) to examine DENSE's structural benefit as far as our workflow permits. In DENSE, Shortcut & Reconcile is tightly coupled to—and attaches directly onto—the subtask tree: as formalized in Section 2.3–2.4 (Equation 1) and Appendix F (Algorithms 1 & 3), it processes trajectories layer by layer from child subtasks to parents, compressing repeated local attempts within each subtask scope and reconciling inherited child issues upward via later recovery evidence. Consequently, once the hierarchical structure is removed, this layer-wise workflow naturally degenerates into a flat whole-trajectory call (w/o hierarchy), which is why our submitted ablation took a single-root form while still isolating structural organization as cleanly as possible.Cross-Recipient
FlatBaseline (Answering W1 & Q1): To directly address your request in Q1, we implementedFlatand evaluated it across all four recipients (3 paired repetitions per task) under the shared REFIT protocol. Operating on the full trajectory under the exact same generator (Gemini 3 Flash Preview), sharedrun0traces, per-call output cap (8,192 tokens), feedback length budget (180/200 KiB), and outcome-blind boundary,Flatperforms two-stage synthesis and evidence review over five structured fields:useful_pathsfailed_approachesissuesresolved,unresolved, orunknown), and the relevant evidence. An issue is marked resolved only when a concrete observation shows recovery. That recovery concerns the specific issue.key_action_indicesrerun_planTable: Strict pass rate (%) across all four recipients (parentheses show
DENSE−Flatin pp).DENSE consistently outperforms Flat across all four recipients (+3.00 to +7.82 pp). Inspecting paired feedback and trajectories confirms two core capabilities of DENSE that
Flatlacks: (i) scope-aware separation of local recovery from task-level completion—without nested subtask boundaries,Flatfrequently marks repaired intermediate sub-steps as[resolved]while overlooking unwritten or unverified final deliverables that DENSE keeps explicitly expanded at higher levels; and (ii) state-guided evidence folding without speculative over-constraint—whereas DENSE folds observation-verified local progress and highlights only open obligations,Flat's unverified globalrerun_plancan omit subtle failure modes or steer recipients into redundant re-exploration.