Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops

Bo Qu; Licheng Wang; Mingguang Chen

REVIEW 3 major objections 8 minor 3 cited by

Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops

T0 review · 3 major / 8 minor · reviewed 2026-07-09 · glm-5.2

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Self-improving AI lives or dies by its verifier

desk verdict A well-executed survey with a genuinely useful taxonomy, whose central empirical claim is honestly caveated but unvalidated read the letter →

arxiv 2607.07663 v1 pith:QNTXHFKR submitted 2026-07-08 cs.AI

Mingguang Chen , Licheng Wang , Bo Qu This is my paper · ORCID

classification cs.AI

keywords recursiveself-improvementverificationhierarchyself-evaluationAIsafetymodelcollapseLLMself-trainingevaluatorco-evolutiongovernance

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This survey of 1,250 papers argues that the entire landscape of AI self-improvement — from inference-time output revision to autonomous research agents — is governed by a single organizing principle: the strength of any self-improvement loop is bounded by the reliability of the signal that tells it whether it got better. The authors propose a two-axis taxonomy separating what a system improves (its outputs, its weights, its evaluator, or the research process itself) from how closed the loop is (human-in-the-loop to fully autonomous). Across all four categories, they observe that demonstrated improvement tracks a verification hierarchy: formal proof checkers at the top, execution feedback next, learned judges below that, and intrinsic self-assessment at the bottom. The characteristic failure modes — self-confirming loops, model collapse, diversity collapse — arise precisely when a system's evaluator sits too low on this hierarchy relative to its ambition. The survey's most consequential claim is that the bottleneck preventing fully autonomous AI research (choosing which problems matter) and the bottleneck preventing reliable self-evaluation are the same bottleneck, and that this bottleneck sits at the top of the verification hierarchy where no current system can operate without human judgment.

What carries the argument

The verification hierarchy: a four-level ordering of evaluator reliability from formal verifiers (strongest) through execution feedback, learned judges, to intrinsic self-assessment (weakest), which the authors claim predicts both where self-improvement works and where it fails.

What would settle it

Find a substantial body of cases where self-improvement loops succeed reliably using only intrinsic signals (model confidence, self-consistency) without external verification — this would break the claimed hierarchy by showing the bottom rung can sustain improvement at scale.

Watch

Extended reading notes

Core claim

The central discovery is the verification hierarchy as an empirical regularity across the self-improvement literature. Formal verifiers (proof checkers, type systems) are sound by construction and permit indefinite iteration without accepting false improvements. Execution feedback (tests, benchmarks) is reliable but incomplete. Learned judges (reward models, LLM-as-judge) are bounded by the judge's own competence and are themselves gameable. Intrinsic signals (confidence, self-consistency) are cheapest and most gameable. The authors observe that every demonstrated self-improvement success in the corpus sits at the top two rungs (code, math, formal verification), while every persistent gap —,

Load-bearing premise

The verification hierarchy is described as a qualitative pattern observed across the corpus, not a measured law. Because the corpus was assembled by seed queries that may systematically over-represent verifiable domains (code, math) where self-improvement works and under-represent domains where weaker signals succeed, the hierarchy could partly reflect sampling bias rather than a genuine empirical regularity about AI self-improvement.

Editorial extensions

If this is right

If the verification hierarchy is real, the path to more autonomous AI research runs through evaluator engineering, not raw model capability — the binding constraint is building systems that can reliably judge open-ended quality, not systems that are merely smarter.
Governance of self-improving AI becomes a measurement problem: regulators would need to audit not what a system produces but what evaluator it runs against and where that evaluator sits on the hierarchy.
The distinction between bounded self-refinement (convergent, evaluable, already industrial) and open-ended RSI (divergent, unverified, still theoretical) gives policymakers a concrete vocabulary for separating near-term engineering risks from speculative takeoff scenarios.
Evaluator co-evolution — the emerging practice of having systems improve their own verifiers alongside their policies — is identified as the pivotal empirical question: it either escapes the self-confirming loop or gives it a second story, and the answer determines whether closed-loop self-improvement stabilizes or compounds bias.
The field's smallest category (foundations, limits, and safety at 60 of 1,250 papers) represents the largest mismatch between stated stakes and research investment, suggesting the academic incentive structure is underweighting exactly the questions that governance depends on.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the verification hierarchy generalizes beyond the corpus, one would predict that self-improvement in domains with cheap verifiers (software, formal math, competitive programming) will continue to advance rapidly while domains without them (scientific research, creative work, strategic planning) will stall regardless of model scale — a domain-specific ceiling on RSI.
The exchange rate of grounding — the minimum fraction of external signal needed to prevent collapse — is a measurable quantity that could be established experimentally by systematically varying the ratio of exogenous to self-generated data in training loops and observing the collapse threshold, turning a theoretical question into an empirical one.
If research taste (identifying which problems are worth working on) is the top rung of the hierarchy and is beginning to acquire a formal theory, then benchmarks for interestingness or problem-selection quality would be the single highest-leverage intervention for the field, as they would convert the last human-in-the-loop bottleneck into a measurable target.
The A-Evolve-Training episode, where a system detected its own proxy metric corrupting and revised its search policy, may represent the first field observation of a system climbing the verification hierarchy autonomously — if this capability generalizes, it would be evidence that evaluator self-correction is achievable, weakening the claim that the verification bottleneck is permanent.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Desk Editor's Note

read the letter

Bottom line: this is a strong organizational survey of a fragmented literature. The two-axis taxonomy (what is improved × loop closure) and the dedicated treatment of self-evaluation as a load-bearing pillar are real contributions. The verification hierarchy — formal verifiers at the top, intrinsic self-assessment at the bottom — is a useful organizing claim, but it is explicitly labeled qualitative and rests on a corpus whose construction was partly shaped by the taxonomy that encodes the hierarchy concept. That circularity is the paper's main soft spot, and the authors are transparent about it, which is appropriate but does not resolve it. The paper deserves a serious referee. The taxonomy and the synthesis of failure modes across sub-fields are the durable contributions. The corpus and classification scripts are released, which is real evidence of methodological transparency. The argumentative skeleton rests on verified anchor works (STaR, Self-Refine, FunSearch, AlphaEvolve, Huang et al.'s negative result on self-correction, Shumailov et al.'s model collapse result) rather than on the 2026 mass, which is the right approach given the recency bias. The within-domain evidence for the hierarchy is genuinely supportive: Huang et al. [27] on intrinsic self-correction failure, the Mirror Loop study [140] on ungrounded self-critique, Lin's rise-and-collapse [92], and the entropy-reservoir unification [145] are all independent work that lines up with the claim. The cross-domain regularity — the claim that makes this a 'reframing of the RSI debate' — is where the circularity concern lands. The supplemental harvest targeted directions 'the taxonomy makes first-class' (§2.3), including self-evaluation as a dedicated category. If every loop is framed as depending on an evaluator, and loops with stronger evaluators are observed to work better, the conclusion is partly built into the lens. The taxonomy elevates self-evaluation, the supplemental harvest fills that category, and then the hierarchy is observed across all categories including the one constructed to embody it. This does not make the claim wrong — the within-domain evidence is independent and supportive — but the cross-domain regularity rests on a corpus whose structure was not independent of the claim being tested. Converting the qualitative observation into a quantitative one (effect sizes by evaluator type across an independently-sampled corpus) is the missing step. The single-annotator classification and the recency bias (74% from 2026) are minor concerns relative to this; the authors flag both honestly. The A-Evolve-Training episode [175] — where a loop detected its own proxy's corruption and revised its search policy — is a striking field observation and deserves the attention the paper gives it. This paper is for researchers working on self-improvement loops, RSI safety, and AI governance who need a map of a rapidly expanding field. It provides one. The framework's durability depends on future quantitative validation of the hierarchy regularity, but the taxonomy and failure-mode synthesis are useful now regardless of whether the hierarchy turns out to be a sampling artifact or a real regularity. Recommend accepting for peer review. The referee should push the authors to (1) acknowledge the framing-observation circularity more explicitly, (2) test whether the hierarchy survives within the seed corpus alone (pre-supplemental harvest), and (3) consider whether any quantitative validation is feasible even at a coarse grain. None of these are blockers for review.

Referee Report

3 major / 8 minor

Summary. This manuscript surveys 1,250 arXiv papers (2024–2026) on AI self-improvement, organizing them along two axes: what the system improves (deployment-time behavior, training-time policy, evaluator, or research process) and the degree of loop closure (human-in-the-loop to fully closed). The central organizing claim is a 'verification hierarchy'—that demonstrated self-improvement strength tracks the reliability of the evaluation signal, from formal verifiers (strongest) to intrinsic self-assessment (weakest)—and that characteristic failure modes (self-confirming loops, model collapse, diversity collapse) follow from violations of this hierarchy. The survey covers four technical categories (deployment-time self-evolution, training-time self-iteration, self-evaluation, auto research) plus a foundations/limits/safety family, and connects the technical literature to RSI theory and governance questions. The corpus, classification scripts, and per-paper assignments are released as reproducible artifacts.

Significance. The paper addresses a genuine gap: no existing survey spans the full spectrum from bounded self-refinement to open-ended RSI while treating self-evaluation as a load-bearing pillar. The two-axis taxonomy is a useful organizing contribution, and the verification hierarchy, if valid, reframes the RSI debate from speculation about capability growth into a measurement program focused on evaluator reliability. The authors ship reproducible corpus construction scripts and per-paper assignments—a concrete strength. The argumentative skeleton rests on verified anchor works (STaR, Self-Refine, FunSearch, Huang et al.'s negative result on intrinsic self-correction) and the diagnostic literature is treated as load-bearing rather than incidental. The connection to Anthropic's autonomy continuum and the identification of governance-grade measurement as an underpopulated niche are well-motivated.

major comments (3)

§5.2 and §5.4: The verification hierarchy is the paper's central load-bearing empirical claim, but it is validated only qualitatively within a corpus whose construction was partly shaped by the taxonomy that encodes the hierarchy. The supplemental harvest (§2.3) targeted 379 papers in 'directions the taxonomy makes first-class,' including self-evaluation methods. The paper's organizing premise—'every improvement loop is a claim that some signal can substitute for human judgment' (§5)—predisposes the analysis to find that evaluator quality is the key variable. If every loop is framed as depending on an evaluator, and loops with stronger evaluators are observed to work better, the conclusion is partly built into the lens. The authors explicitly caveat this ('a qualitative pattern we observe throughout the corpus, not a measured law'), but the caveat does not resolve the structural concern.
§5.4: The claim that 'the direction-setting bottleneck and the verification bottleneck are the same bottleneck' is asserted rather than demonstrated. The Anthropic essay [5] identifies research direction-setting as a distinct bottleneck (choosing which problems matter), while the verification hierarchy concerns signal reliability for already-specified tasks. The conflation of these two different bottlenecks—problem selection and solution verification—is load-bearing for the paper's framing but is not separately argued. The within-domain evidence (Huang et al. [27], Mirror Loop [140], Lin [92]) supports the verification claim but does not address whether direction-setting reduces to verification.
§2.3: The corpus construction has a recency bias (74% from 2026) that interacts with the verification hierarchy claim in a way that is not fully addressed. If the field has recently concentrated on verifiable domains (code, math) where self-improvement works, the hierarchy could partly reflect where current research activity happens to be dense rather than a stable empirical regularity. The authors acknowledge this for citation counts but not for the hierarchy claim itself. A simple robustness check—reporting the hierarchy's support stratified by publication year—would help distinguish a real regularity from a recency artifact.

minor comments (8)

Figure 1 is referenced as laying out the 4×3 grid with representative systems, but the figure content is not visible in the manuscript text. Ensure the figure is properly rendered and legible in the final version.
Figure 2 (semantic map) uses TF-IDF abstracts with SVD + t-SNE projection. The axes are described as 'arbitrary embedding dimensions' but no clustering quality metric (e.g., silhouette score) is reported to support the claim that 'Auto Research and foundations form coherent regions.'
Table 1 reports '82% posted in 2026' for the self-evaluation category, which is notably higher than other categories (57–76%). Given that the supplemental harvest specifically targeted self-evaluation methods, this percentage may reflect sampling design rather than field dynamics. A footnote clarifying this would help.
§4.3: The claim that on-policy self-distillation 'did not exist before 2026' is strong; the authors should verify whether precursor work (e.g., earlier self-distillation variants) existed under different names.
§7.1: The Whitfill and Wu result [189] is described as 'the empirical crux of the RSI-feasibility debate,' but the two specifications diverging (substitutes vs. complements) means the result is inconclusive rather than a crux. The framing should be softened.
The reference list includes several papers from 2026 with arXiv IDs starting '2607' (July 2026), which is the same month as this manuscript's submission. The authors should verify these are not concurrent submissions that could create citation circularity.
§3.5: The term 'harness' is defined in §2.1 but used extensively before its definition appears in the reading order. Consider forward-referencing the definition in §3.
The abstract states '1,250 arXiv papers' while §2.3 describes 871 seed + 379 supplemental = 1,250. This is consistent but the abstract could note the two-stage construction for clarity.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for a careful and constructive report. The referee identifies three major concerns: (1) a potential circularity in the verification hierarchy claim arising from the taxonomy-guided corpus construction, (2) an insufficiently argued conflation of the direction-setting bottleneck with the verification bottleneck, and (3) an unaddressed interaction between corpus recency bias and the hierarchy claim. We agree that all three points identify genuine weaknesses in the current manuscript and will revise accordingly. Specifically, we will (a) add an explicit methodological caveat about the taxonomy-corpus interaction and reframe the hierarchy's epistemic status, (b) separate the two bottlenecks and argue the reduction rather than asserting it, and (c) add a stratified robustness check by publication year. We cannot fully resolve the structural circularity concern—it is inherent to any taxonomy-guided survey—but we can make it fully transparent and distinguish what the hierarchy claims to be (an organizing empirical pattern) from what it does not claim to be (an independently validated law).

read point-by-point responses

Referee: §5.2 and §5.4: The verification hierarchy is the paper's central load-bearing empirical claim, but it is validated only qualitatively within a corpus whose construction was partly shaped by the taxonomy that encodes the hierarchy. The supplemental harvest (§2.3) targeted 379 papers in 'directions the taxonomy makes first-class,' including self-evaluation methods. The paper's organizing premise—'every improvement loop is a claim that some signal can substitute for human judgment' (§5)—predisposes the analysis to find that evaluator quality is the key variable. If every loop is framed as depending on an evaluator, and loops with stronger evaluators are observed to work better, the conclusion is partly built into the lens.

Authors: The referee is correct that there is a structural circularity here, and we had not stated it with sufficient force. The concern has two layers, and we address each. First, the corpus-construction layer: the supplemental harvest (§2.3) targeted self-evaluation methods, test-time training, and zero-data self-play because the taxonomy made them first-class. This means the corpus is not an independent sample with respect to the taxonomy—it was partly shaped by it. We will add an explicit methodological caveat in §2.3 stating this directly and noting that the corpus proportions cannot be read as unbiased estimates of the field's composition. Second, the conceptual layer: the premise that 'every improvement loop is a claim that some signal can substitute for human judgment' does frame every loop as evaluator-dependent, which makes the finding that evaluator quality matters partly built into the lens. We acknowledge this honestly. However, we note that the premise is not merely an assumption—it is a logical observation: any improvement loop requires a criterion for 'better,' and that criterion is an evaluator by definition. What is not built into the lens is the empirical claim that stronger evaluators (formal verifiers) produce more durable improvement than weaker ones (intrinsic self-assessment). That claim is supported by independent anchor evidence (Huang et al. [27], the Mirror Loop [140], Lin [92], FunSearch [3], AlphaEvolve [4]) that was not selected by the taxonomy but by its diagnostic quality. Still, the referee is right that the qualitative pattern across the full corpus is observed through the taxonomy's lens, and we will reframe the hierarchy's epistemic status accordingly. We will (a) add a paragraph in §5.2 explicitly acknowledging the taxonomy-corpus circularly revision: partial
Referee: §5.4: The claim that 'the direction-setting bottleneck and the verification bottleneck are the same bottleneck' is asserted rather than demonstrated. The Anthropic essay [5] identifies research direction-setting as a distinct bottleneck (choosing which problems matter), while the verification hierarchy concerns signal reliability for already-specified tasks. The conflation of these two different bottlenecks—problem selection and solution verification—is load-bearing for the paper's framing but is not separately argued. The within-domain evidence (Huang et al. [27], Mirror Loop [140], Lin [92]) supports the verification claim but does not address whether direction-setting reduces to verification.

Authors: The referee is correct. The claim as stated conflates two distinct bottlenecks: (1) verifying that a solution to a specified problem is correct, and (2) selecting which problems are worth solving. The Anthropic essay [5] treats these as distinct, and our evidence base (Huang et al. [27], Mirror Loop [140], Lin [92]) supports only the first claim. We will revise §5.4 to separate the two bottlenecks and argue the relationship rather than asserting identity. The argument we can honestly make is narrower: direction-setting is a harder problem than verification, and it sits at or above the top of the verification hierarchy because it requires evaluating not just whether a solution is correct but whether a problem is worth posing—a judgment for which no formal verifier exists. This makes direction-setting a superset of the verification problem, not the same problem. The evidence for this narrower claim comes from §6.3's diagnostic literature (ScienceAgentBench [177], ResearchArena [178], the integrity benchmarks [143]) and from Herrmann and Schmidhuber's formalization of interestingness as a bottleneck [190], which identifies problem selection as requiring inductive heuristics that go beyond verification. We will rewrite the claim to state that direction-setting subsumes the verification bottleneck and adds the further difficulty of evaluating problem importance, for which no trustworthy automated signal currently exists. We will remove the assertion of identity and replace it with this argued relationship. revision: yes
Referee: §2.3: The corpus construction has a recency bias (74% from 2026) that interacts with the verification hierarchy claim in a way that is not fully addressed. If the field has recently concentrated on verifiable domains (code, math) where self-improvement works, the hierarchy could partly reflect where current research activity happens to be dense rather than a stable empirical regularity. The authors acknowledge this for citation counts but not for the hierarchy claim itself. A simple robustness check—reporting the hierarchy's support stratified by publication year—would help distinguish a real regularity from a recency artifact.

Authors: This is a fair and actionable suggestion. We will add the requested stratified robustness check: for each rung of the verification hierarchy, we will report the distribution of supporting evidence across publication years (2024, 2025, 2026). This will let readers assess whether the hierarchy is driven by the 2026 mass or holds across the corpus's temporal span. We expect the hierarchy to be visible in the 2024 anchor works (STaR [7], Self-Refine [6], Huang et al. [27], FunSearch [3]) and in the 2025 diagnostic literature (Mirror Loop [140], Lin [92]), not only in the 2026 wave, but the referee is right that this should be shown rather than assumed. We will also add an explicit caveat in §5.2 noting that the hierarchy's support is concentrated in verifiable domains (code, math, formal methods) and that its applicability to non-verifiable domains is inferred by absence (the lack of demonstrated durable self-improvement without external signal) rather than by positive evidence of failure. The recency interaction is real: if the field's attention has shifted toward verifiable domains, the hierarchy could partly reflect where activity is dense. The stratified check will help distinguish these possibilities, though we acknowledge it cannot fully separate a genuine regularity from a field-wide attention shift—both could produce the same stratified pattern. We will state this limitation explicitly. revision: yes

standing simulated objections not resolved

The structural circularity between taxonomy-guided corpus construction and the verification hierarchy claim cannot be fully resolved. The taxonomy shapes the corpus, and the corpus is used to validate a claim encoded in the taxonomy. We can make this fully transparent, reframe the hierarchy's epistemic status as a qualitative pattern rather than a measured law, and point to independent anchor evidence, but we cannot make the corpus independent of the taxonomy without rebuilding it from scratch using taxonomy-free queries—which would lose the targeted coverage of self-evaluation that makes the survey distinctive.

Circularity Check

2 steps flagged · score 2.0 of 10

Mild self-definitional framing (improvement defined via evaluator → evaluator quality is the bottleneck), but transparently acknowledged and externally grounded; no self-citation chain.

self definitional [§2.1 (definition of Self-improvement) and §5 (verification hierarchy claim)]
"Self-improvement. A system participates in producing a better version of itself or of its own outputs, where “better” is defined by some evaluator. The definitional dependence on an evaluator is not pedantry; it is the source of every failure mode in §5."

The paper defines self-improvement as inherently dependent on an evaluator ('better is defined by some evaluator'), then derives that evaluator quality is the universal bottleneck ('the loop's ceiling is exactly the quality of that substitute'). If 'improvement' is defined as 'improvement against an evaluator,' then 'evaluator quality bounds improvement quality' follows partly by definition. However, the paper is explicitly transparent about this ('not pedantry; it is the source of every failure mode'), uses it as a deliberate framing device rather than a hidden derivation, and the specific empirical content—the ordering of the hierarchy (formal verifiers > execution > learned judges > intrinsic signals) and the three failure modes—is supported by independent external work (Huang et al. [2
fitted input called prediction [§2.3 (corpus construction) and §5.2 (hierarchy observation)]
"a targeted supplemental harvest of 379 papers in directions the taxonomy makes first-class: self-evaluation methods (judges, process reward models, verifiers, rubrics, meta-evaluation), test-time training, and zero-data self-play. [...] The empirical regularity across all four categories—a qualitative pattern we observe throughout the corpus, not a measured law—is that demonstrated self-improvement strength tracks this hierarchy."

The taxonomy elevates self-evaluation to a first-class category, the supplemental harvest fills that category with 318 papers, and then the verification hierarchy is observed across all categories including the one constructed to embody it. This creates a mild framing-observation loop: the lens (taxonomy) partly shapes the data (supplemental harvest), and the conclusion (hierarchy) is read from data shaped by the lens. However, the paper is transparent about this ('the raw counts partly mirror our own sampling'), explicitly labels the hierarchy as 'not a measured law,' and the within-domain evidence supporting the hierarchy (Huang et al. [27], Lin [92], Mirror Loop [140], Shumailov et al. [144]) is drawn from independent external work, not from the corpus construction itself. The circulari

full rationale

This is a survey paper, not a formal derivation, so the circularity patterns targeting fitted-parameter predictions or self-citation chains largely do not apply. No self-citations exist in the reference list (authors Chen, Wang, Qu cite none of their own prior work), ruling out patterns 3–5. The two mild concerns are: (1) a self-definitional framing where 'improvement' is defined via an evaluator and then evaluator quality is found to be the bottleneck—transparently acknowledged by the paper as a deliberate framing choice; and (2) a corpus-construction loop where the taxonomy shapes the supplemental harvest which then supports the hierarchy observed across categories—transparently acknowledged as 'not a measured law' with corpus proportions explicitly noted to 'partly mirror our own sampling.' In both cases the paper is candid about the limitation, and the specific empirical claims (hierarchy ordering, failure modes) are grounded in independent external citations. The circularity is real but minor and self-disclosed, warranting a score of 2.

Assumptions & free parameters 3 free parameters · 4 assumptions · 2 invented entities

The survey introduces no new physical entities or fitted constants. The free parameters are methodological choices (query caps, harvest targets, classification rules) that shape the corpus. The axioms are domain assumptions about the nature of self-improvement and evaluator reliability, all stated explicitly and reasonable within the field. The invented entities are organizational constructs (the hierarchy, the taxonomy) rather than postulated physical objects, and both have falsifiable handles.

free parameters (3)

seed query depth caps = not specified
Per-query caps on arXiv harvest affect corpus composition; the authors acknowledge these favor recent, high-volume threads.
supplemental harvest target directions = self-evaluation, test-time training, zero-data self-play
The three directions chosen for supplemental harvest were determined by the taxonomy, creating a feedback between the framework and the corpus.
keyword classification rules = not fully specified in text
Rule-based classification moved 89 papers off thread defaults; the rules themselves are in released scripts but their design is a free parameter of the method.

assumptions (4)

domain assumption Every self-improvement loop depends on an evaluator that defines 'better.'
Stated in §2.1 ('Self-improvement' definition) and §5 opening. This is the foundational premise of the survey's argument. It is a reasonable domain assumption but is not independently proven; it is asserted as definitional.
domain assumption The verification hierarchy (formal verifiers > execution feedback > learned judges > intrinsic signals) is a meaningful ordering of evaluator reliability.
Introduced in §5.2. The ordering is intuitive and supported by cited examples, but the paper explicitly calls the tracking claim 'a qualitative pattern we observe throughout the corpus, not a measured law.' The hierarchy itself is treated as an axiom for organizing the survey.
domain assumption Bounded self-refinement and open-ended RSI are categorically distinct phenomena.
Stated in §2.1 ('Bounded self-refinement vs. open-ended RSI'). This distinction is the survey's 'central cut' and is assumed rather than derived. The theory literature in §7 is cited as support, but the categorical distinction is an organizing premise.
domain assumption The arXiv corpus is a representative sample of the self-improvement literature.
Implicit in the survey methodology (§2.3). The authors acknowledge the corpus is 'a sample, not a census' and note publication-censoring effects for industrial RSI practice, but the survey's claims depend on the sample being informative about the field's structure.

invented entities (2)

Verification hierarchy (4-level) independent evidence
purpose: Orders evaluator signals from formal verifiers (strongest) to intrinsic self-assessment (weakest) to explain why some self-improvement loops work and others fail.
Not a new physical entity but a new organizational construct. It is supported by cited empirical evidence (Huang et al. [27], Lightman et al. [130], Shumailov et al. [144]) and makes a falsifiable prediction: systems using higher-rung verifiers will show stronger, more stable self-improvement. The prediction is qualitative and not yet quantitatively tested in the paper.
Two-axis taxonomy (4 categories × 3 closure levels) independent evidence
purpose: Classifies self-improvement methods by what they improve and who validates the improvement.
An organizational framework, not a physical entity. It is validated by the corpus mapping (1,250 papers classified) and makes the field's structure visible. The taxonomy absorbed two paradigm arrivals (OPSD, zero-data self-play) without modification, which the authors cite as evidence of durability.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops." pith.science (2026). https://pith.science/paper/QNTXHFKR

@misc{pith2026260707663,
  author       = {Pith},
  title        = {Pith review of: Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QNTXHFKR}},
  note         = {Machine review of arXiv:2607.07663}
}

read the original abstract

AI systems increasingly participate in their own improvement: revising their outputs, adapting their own harnesses during deployment, training on data they generate, and, increasingly, conducting AI research itself. This literature is described under a vocabulary ("self-refine," "self-reward," "self-play," "self-evolve") that conflates fundamentally different ambitions. We survey 1,250 arXiv papers (2024-2026) along two axes: what the system improves -- its behavior in deployment, its policy through training, its evaluator, or the research process itself -- and the degree of loop closure (human-in-the-loop to fully closed). The taxonomy separates bounded self-refinement -- convergent, evaluable, and already industrial practice -- from open-ended recursive self-improvement (RSI), which remains bounded by grounding requirements, collapse dynamics, and compute constraints on every measured axis. Its distinctive feature is a dedicated category for self-evaluation: every improvement loop is a claim that some signal can substitute for human judgment. We survey the evaluator design space -- judges, process reward models, verifiers, rubrics, meta-evaluation -- order the signals into a verification hierarchy from formal verifiers (strongest) to intrinsic self-assessment (weakest), and observe that demonstrated self-improvement strength tracks this hierarchy, that its failure modes (self-confirming loops, model collapse, diversity collapse) follow from its violations, and that the "research direction-setting" bottleneck keeping humans in the loop sits at the top of that hierarchy. We connect the technical literature to the theory of RSI limits and to the safety and governance questions raised by frontier-lab accounts of closing the loop, and identify governance-grade measurement of self-improvement as the field's most underpopulated niche.

Figures

Figures reproduced from arXiv: 2607.07663 by the authors.

**Figure 2.** semantic map of the 1,250-paper corpus (TF-IDF abstracts, SVD + t-SNE projection; [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** deployment-time self-evolution ordered by persistence — refined outputs evaporate with [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: the training-time self-iteration loop. The five paradigms of §4 differ mainly in who [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]

**Figure 5.** Figure 5: the verification hierarchy. Signal reliability rises toward the top while task coverage [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]

**Figure 6.** Figure 6: Seed-corpus quarterly growth through 2026Q2, showing log-scaled paper counts and [PITH_FULL_IMAGE:figures/full_fig_p026_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

The Calibration Floor: Format Repair Can Masquerade as Self-Correction at Small-to-Mid Scale
cs.CL 2026-08 conditional novelty 7.0 of 10

Across 0.8B-12B models plus a frontier arm, apparent self-correction effects are dominated by format-recovery and format-loss artifacts, with near-zero content-level change at capable scale.
Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA
cs.LG 2026-08 conditional novelty 6.0 of 10

The released Macaron-V1-Venti model uses a frozen 744B base plus four per-turn-routed LoRA specialists and reports high internal benchmark scores, but it does not demonstrate cross-generation continual-learning gains.
The Horizon Gap: Planning, Memory, Execution, Training, and Evaluation for Long-Horizon LLM Agents
cs.CL 2026-08 conditional novelty 5.0 of 10

A survey of 1,547 papers defines the 'horizon gap' and documents that long-horizon agent research is converging on trajectory-level process signals instead of outcome-only scores.

Reference graph

Works this paper leans on

205 extracted references · 205 canonical work pages · cited by 3 Pith papers

[27]

Adaptive self-improvement LLM agentic system for ML library development, 2025

Genghan Zhang, Weixin Liang, Olivia Hsu, and Kunle Olukotun. Adaptive self-improvement LLM agentic system for ML library development, 2025. URL https://arxiv.org/abs/2502.0 2534. 30

work page 2025
[140]

Efficient Process Reward Modeling via Contrastive Mutual Information

Nakyung Lee, Sangwoo Hong, and Jungwoo Lee. Eﬀicient process reward modeling via contrastive mutual information, 2026. URL https://arxiv.org/abs/2604.10660

work page Pith review arXiv 2026
[92]

Retrospective progress-aware self-refinement for LLM agent training, 2026

Xinbei Ma, Congmin Zheng, Jiyang Qiu, Jiale Hong, Yao Yao, Xiangmou Qu, Jiaxin Yin, Xingyu Lou, Jun Wang, Weiwen Liu, Weinan Zhang, Zhuosheng Zhang, and Hongyun Zhao. Retrospective progress-aware self-refinement for LLM agent training, 2026. URL https://ar xiv.org/abs/2606.14302

work page arXiv 2026
[145]

Ali Heydari, Simon A

Weizhi Zhang, Zechen Li, Hamid Palangi, Ben Graef, A. Ali Heydari, Simon A. Lee, Sabera Rahman, Ray Luo, Zeinab Esmaeilpour, Erik Schenck, Chloe Zhang, Yamin Li, Menglian Zhou, Philip S. Yu, Daniel McDuff, Lindsey Sunden, Mark Malhotra, Shwetak Patel, and Ahmed A. Metwally. RubricsTree: Scalable and evolving open-ended evaluation of personal health agents...

work page 2026
[175]

Effective Harness Engineering for Algorithm Discovery with Coding Agents

Yoichi Ishibashi, Taro Yano, and Masafumi Oyamada. Effective harness engineering for algorithm discovery with coding agents, 2026. URL https://arxiv.org/abs/2605.15221

work page Pith review arXiv 2026
[5]

Recursive self-improvement

Anthropic. Recursive self-improvement. Anthropic Institute blog post, 2026. URL https: //www.anthropic.com/institute/recursive-self-improvement. Published May 2026

work page 2026
[1]

Speculations concerning the first ultraintelligent machine

Irving John Good. Speculations concerning the first ultraintelligent machine. Advances in Computers, 6:31–88, 1966. doi: 10.1016/S0065-2458(08)60418-0

work page doi:10.1016/s0065-2458(08)60418-0 1966
[2]

Goedel Machines: Self-Referential Universal Problem Solvers Making Provably Optimal Self-Improvements

Jürgen Schmidhuber. Gödel machines: Self-referential universal problem solvers making provably optimal self-improvements. arXiv preprint cs/0309048 , 2003. URL https://arxiv.or g/abs/cs/0309048

work page Pith review arXiv 2003

Show all 205 references

[3]

Pawan Kumar, Emilien Dupont, Francisco J

Bernardino Romera-Paredes, Mohammadamin Barekatain, Alexander Novikov, Matej Balog, M. Pawan Kumar, Emilien Dupont, Francisco J. R. Ruiz, Jordan S. Ellenberg, Pengming Wang, Omar Fawzi, Pushmeet Kohli, and Alhussein Fawzi. Mathematical discoveries from program search with larg...

2024 doi
[4]

Alexander Novikov, Ngân V u, Marvin Eisenberger, Emilien Dupont, Po-Sen Huang, Adam Zsolt Wagner, Sergey Shirobokov, Borislav Kozlovskii, Francisco J. R. Ruiz, Abbas Mehrabian, M. Pawan Kumar, Abigail See, Swarat Chaudhuri, George Holland, Alex Davies, Sebastian Nowozin, Pushm...

2025 arXiv
[6]

Self-refine: Iterative refinement with self-feedback

Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegr- effe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, Shashank Gupta, Bod- hisattwa Prasad Majumder, Katherine Hermann, Sean Welleck, Amir Yazdanbakhsh, and Peter Clark. Self-refine: Iter...

2023 arXiv
[7]

Eric Zelikman, Yuhuai Wu, Jesse Mu, and Noah D. Goodman. STaR: Bootstrapping reasoning with reasoning. In Advances in Neural Information Processing Systems (NeurIPS), 2022. URL https://arxiv.org/abs/2203.14465

2022 arXiv
[8]

Self-rewarding language models

Weizhe Yuan, Richard Yuanzhe Pang, Kyunghyun Cho, Xian Li, Sainbayar Sukhbaatar, Jing Xu, and Jason Weston. Self-rewarding language models. In International Conference on Machine Learning (ICML) , 2024. URL https://arxiv.org/abs/2401.10020

2024 arXiv
[9]

Gödel agent: A self-referential agent framework for recursive self-improvement, 2024

Xunjian Yin, Xinyi Wang, Liangming Pan, Li Lin, Xiaojun Wan, and William Yang Wang. Gödel agent: A self-referential agent framework for recursive self-improvement, 2024. URL https://arxiv.org/abs/2410.04444

2024 arXiv
[10]

Darwin gödel machine: Open-ended evolution of self-improving agents, 2025

Jenny Zhang, Shengran Hu, Cong Lu, Robert Lange, and Jeff Clune. Darwin gödel machine: Open-ended evolution of self-improving agents, 2025. URL https://arxiv.org/abs/2505.22954

2025 arXiv
[11]

The AI scientist: Towards fully automated open-ended scientific discovery, 2024

Chris Lu, Cong Lu, Robert Tjarko Lange, Jakob Foerster, Jeff Clune, and David Ha. The AI scientist: Towards fully automated open-ended scientific discovery, 2024. URL https: //arxiv.org/abs/2408.06292

2024 arXiv
[12]

A survey of on-policy distillation for large language models,

Mingyang Song and Mao Zheng. A survey of on-policy distillation for large language models,
[13]

URL https://arxiv.org/abs/2604.00626. 29

arXiv
[14]

Jiaqi Wei, Xiang Zhang, Yuejin Yang, Wenxuan Huang, Juntai Cao, Sheng Xu, Xiang Zhuang, Zhangyang Gao, Muhammad Abdul-Mageed, Laks V. S. Lakshmanan, Chenyu You, Wanli Ouyang, and Siqi Sun. Unifying tree search algorithm and reward design for LLM reasoning: A survey, 2025. URL ...

2025
[15]

Stop hand-holding your coding agent: Engineering the loops that replace step-by-step prompting, 2026

Sandeco Macedo. Stop hand-holding your coding agent: Engineering the loops that replace step-by-step prompting, 2026. URL https://arxiv.org/abs/2607.00038

2026 arXiv
[16]

Absolute zero: Reinforced self-play reasoning with zero data, 2025

Andrew Zhao, Yiran Wu, Yang Yue, Tong Wu, Quentin Xu, Yang Yue, Matthieu Lin, Shenzhi Wang, Qingyun Wu, Zilong Zheng, and Gao Huang. Absolute zero: Reinforced self-play reasoning with zero data, 2025. URL https://arxiv.org/abs/2505.03335

2025 arXiv
[17]

R-zero: Self-evolving reasoning LLM from zero data,

Chengsong Huang, Wenhao Yu, Xiaoyang Wang, Hongming Zhang, Zongxia Li, Ruosen Li, Jiaxin Huang, Haitao Mi, and Dong Yu. R-zero: Self-evolving reasoning LLM from zero data,
[18]

ICLR 2026

URL https://arxiv.org/abs/2508.05004. ICLR 2026

2026 arXiv
[19]

Reflexion: Language agents with verbal reinforcement learning

Noah Shinn, Federico Cassano, Edward Berman, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. Reflexion: Language agents with verbal reinforcement learning. In Advances in Neural Information Processing Systems (NeurIPS) , 2023. URL https://arxiv.org/abs/2303.1 1366

2023
[20]

Scaling LLM test-time compute optimally can be more effective than scaling model parameters, 2024

Charlie Snell, Jae-Hoon Lee, Kelvin Xu, and A viral Kumar. Scaling LLM test-time compute optimally can be more effective than scaling model parameters, 2024. URL https://arxiv.or g/abs/2408.03314

2024 arXiv
[21]

SymbolicAI: A framework for logic-based approaches combining gener- ative models and solvers, 2024

Marius-Constantin Dinu, Claudiu Leoveanu-Condrei, Markus Holzleitner, Werner Zellinger, and Sepp Hochreiter. SymbolicAI: A framework for logic-based approaches combining gener- ative models and solvers, 2024. URL https://arxiv.org/abs/2402.00854

2024 arXiv
[22]

To- wards reliable and robust LLM planning: Symbolic feedback-driven iterative self-refinement framework, 2026

Jiajing Zhang, Jiamei Jiang, Chenyang Zhang, Feifei Mo, Linjing Li, and Daniel Zeng. To- wards reliable and robust LLM planning: Symbolic feedback-driven iterative self-refinement framework, 2026. URL https://arxiv.org/abs/2606.27757

2026 arXiv
[23]

SQL-o1: A self-reward heuristic dynamic search method for text- to-SQL, 2025

Shuai Lyu, Haoran Luo, Ripeng Li, Zhonghong Ou, Jiangfeng Sun, Yang Qin, Xiaoran Shang, Meina Song, and Yifan Zhu. SQL-o1: A self-reward heuristic dynamic search method for text- to-SQL, 2025. URL https://arxiv.org/abs/2502.11741

2025 arXiv
[24]

Hallucination detection-guided preference optimization for clinical summarization, 2026

Shamanth Kuthpadi Seethakantha, Dung Thai, Vara Prasad Gudi, Simran Tiwari, Rami Matar, A vijit Mitra, Wenlong Zhao, Wael Salloum, and Andrew McCallum. Hallucination detection-guided preference optimization for clinical summarization, 2026. URL https://arxi v.org/abs/2605.28910

2026 arXiv
[25]

LongSumEval: Question-answering based evaluation and feedback-driven refinement for long document sum- marization, 2026

Huyen Nguyen, Haoxuan Zhang, Yang Zhang, Haihua Chen, and Junhua Ding. LongSumEval: Question-answering based evaluation and feedback-driven refinement for long document sum- marization, 2026. URL https://arxiv.org/abs/2604.25130

2026 arXiv
[26]

Albrecht, Peter Bell, and Amos Storkey

Dongge Han, Trevor McInroe, Adam Jelley, Stefano V. Albrecht, Peter Bell, and Amos Storkey. LLM-personalize: Aligning LLM planners with human preferences via reinforced self-training for housekeeping robots, 2024. URL https://arxiv.org/abs/2404.14285

2024 arXiv
[28]

What does LLM refinement actually improve? a systematic study on document-level literary translation, 2026

Shaomu Tan, Dawei Zhu, Ke Tran, Michael Denkowski, Sony Trenous, Bill Byrne, Leonardo Ribeiro, and Felix Hieber. What does LLM refinement actually improve? a systematic study on document-level literary translation, 2026. URL https://arxiv.org/abs/2605.13368

2026 arXiv
[29]

Large language models cannot self-correct reasoning yet, 2024

Jie Huang, Xinyun Chen, Swaroop Mishra, Huaixiu Steven Zheng, Adams Wei Yu, Xinying Song, and Denny Zhou. Large language models cannot self-correct reasoning yet, 2024. URL https://arxiv.org/abs/2310.01798. ICLR 2024

2024 arXiv
[30]

When does intrinsic self- correction help? a task-sensitive analysis, 2026

Elroy Stav, Dvir Berlowitz, Maayan Orner, and Sarit Kraus. When does intrinsic self- correction help? a task-sensitive analysis, 2026. URL https://arxiv.org/abs/2606.23196

2026 arXiv
[31]

When are likely answers right? on sequence probability and correctness in LLMs, 2026

Johannes Zenn and Jonas Geiping. When are likely answers right? on sequence probability and correctness in LLMs, 2026. URL https://arxiv.org/abs/2606.27359

2026 arXiv
[32]

Ask, don’t judge: Binary questions for interpretable LLM evaluation and self-improvement, 2026

Sangwoo Cho, Kushal Chawla, Pengshan Cai, Zefang Liu, Chenyang Zhu, Shi-Xiong Zhang, and Sambit Sahu. Ask, don’t judge: Binary questions for interpretable LLM evaluation and self-improvement, 2026. URL https://arxiv.org/abs/2606.27226

2026 arXiv
[33]

Ll3m: Large language 3d modelers, 2025

Shuang Lu, Chen Guan, Nam Anh Dinh, Itai Lang, Ari Holtzman, and Rana Hanocka. Ll3m: Large language 3d modelers, 2025. URL https://arxiv.org/abs/2508.08228

2025 arXiv
[34]

AgenticDB: Agentic performance reconfiguration for database workloads, 2026

Xiaomin Yang, Chaozheng Wang, Chen Zheng, Heng Zhang, and Yanjun Wu. AgenticDB: Agentic performance reconfiguration for database workloads, 2026. URL https://arxiv.org/ abs/2606.20318

2026 arXiv
[35]

AutoPass: Evidence-guided LLM agents for compiler performance tuning, 2026

Zepeng Li, Jie Ren, Zhanyong Tang, Jie Zheng, and Zheng Wang. AutoPass: Evidence-guided LLM agents for compiler performance tuning, 2026. URL https://arxiv.org/abs/2606.20373

2026 arXiv
[36]

LEAP: Supercharging LLMs for formal mathematics with agentic frameworks, 2026

Po-Nien Kung, Linfeng Song, Dawsen Hwang, Jinsung Yoon, Chun-Liang Li, Simone Sev- erini, Mirek Olšák, Edward Lockhart, Quoc V Le, Burak Göktürk, Thang Luong, Tomas Pfister, and Nanyun Peng. LEAP: Supercharging LLMs for formal mathematics with agentic frameworks, 2026. URL htt...

2026 arXiv
[37]

KVerus: Scalable and resilient formal verification proof generation for rust code, 2026

Yuwei Liu, Xinyi Wan, Yanhao Wang, Minghua Wang, Lin Huang, and Tao Wei. KVerus: Scalable and resilient formal verification proof generation for rust code, 2026. URL https: //arxiv.org/abs/2605.03822

2026 arXiv
[38]

KBSpec: LLM-driven formal specification generation with evolving domain knowledge base, 2026

Wenhan Wang and Zeyu Sun. KBSpec: LLM-driven formal specification generation with evolving domain knowledge base, 2026. URL https://arxiv.org/abs/2606.21339

2026 arXiv
[39]

Verifier-guided code translation via meta-step decoding, 2026

Tianyang Zhou, Somesh Jha, Mihai Christodorescu, Kirill Levchenko, and Varun Chan- drasekaran. Verifier-guided code translation via meta-step decoding, 2026. URL https: //arxiv.org/abs/2605.17626

2026 arXiv
[40]

What drives interactive improvement from feedback?, 2026

Bartłomiej Cupiał, Jan Lojek, Mikołaj Garstecki, Szymon Pobłocki, Alicja Ziarko, and Piotr Miłoś. What drives interactive improvement from feedback?, 2026. URL https://arxiv.org/ abs/2606.30774

2026 arXiv
[41]

Falsification, not exposure: An internally preregistered placebo-controlled decomposition of self-repair feedback in frozen small code models, 2026

Mehmet İşcan. Falsification, not exposure: An internally preregistered placebo-controlled decomposition of self-repair feedback in frozen small code models, 2026. URL https://arxiv. org/abs/2606.31511

2026 arXiv
[42]

Feedback over form: Why execution feedback matters more than pipeline topology in 1-3b code generation, 2026

Charles Junichi McAndrews. Feedback over form: Why execution feedback matters more than pipeline topology in 1-3b code generation, 2026. URL https://arxiv.org/abs/2604.21950. 31

2026 arXiv
[43]

RubricRefine: Improving tool-use agent reliability with training-free pre-execution refinement, 2026

Will LeVine, Brendan Evers, Sam Saltwick, and Abhay Venkatesh. RubricRefine: Improving tool-use agent reliability with training-free pre-execution refinement, 2026. URL https://ar xiv.org/abs/2605.09730

2026 arXiv
[44]

Unlocking LLM code correction with iterative feedback loops,

Le Zhang and Suresh Kothari. Unlocking LLM code correction with iterative feedback loops,
[45]

URL https://arxiv.org/abs/2606.17514

arXiv
[46]

Denoising iterative self-correction: Structured verification loops for reliable LLM reasoning, 2026

Shen Yin, David Ken, and Joel Stremmel. Denoising iterative self-correction: Structured verification loops for reliable LLM reasoning, 2026. URL https://arxiv.org/abs/2606.21724

2026 arXiv
[47]

FLARE: Fine-grained diagnostic feedback for LLM code refinement, 2026

Yinsheng Yao, Hongxiang Zhang, Weixi Tong, and Tianyi Zhang. FLARE: Fine-grained diagnostic feedback for LLM code refinement, 2026. URL https://arxiv.org/abs/2606.03852

2026 arXiv
[48]

CoSPlay: Cooperative self-play at test-time with self- generated code and unit test, 2026

Zhangyi Hu, Chenhui Liu, Tian Huang, Jindong Li, Yang Yang, Jiemin Wu, Zining Zhong, Menglin Yang, and Yutao Yue. CoSPlay: Cooperative self-play at test-time with self- generated code and unit test, 2026. URL https://arxiv.org/abs/2605.23491

2026 arXiv
[49]

Kestrel: Grounding self-refinement for L VLM hallucination mitigation, 2026

Jiawei Mao, Hardy Chen, Haoqin Tu, Yuhan Wang, Letian Zhang, Zeyu Zheng, Huaxiu Yao, Zirui Wang, Cihang Xie, and Yuyin Zhou. Kestrel: Grounding self-refinement for L VLM hallucination mitigation, 2026. URL https://arxiv.org/abs/2603.16664

2026
[50]

Reflect-r1: Evidence-driven reflection for self- correction in long video understanding, 2026

Shuimu Chen, Yihan Chen, Yuanshen Guan, Zebang Cheng, Zeyu Zhang, Su-Juan Qin, Bin Xia, Jiaran Li, Wenming Yang, and Fei Ma. Reflect-r1: Evidence-driven reflection for self- correction in long video understanding, 2026. URL https://arxiv.org/abs/2606.27922

2026 arXiv
[51]

FiRe: Fine-grained multimodal reasoning for enhanced image generation, 2026

Yong-Jin Kim, Yoonjin Oh, Yerin Kim, Hyomin Kim, Jeeyoung Yun, Yujung Heo, Minjun Kim, and Sungwoong Kim. FiRe: Fine-grained multimodal reasoning for enhanced image generation, 2026. URL https://arxiv.org/abs/2604.13491

2026 arXiv
[52]

Proprio: Latent self- scoring and inference-time refinement for physically plausible video generation, 2026

Mariam Hassan, Kaouther Messaoud, Wuyang Li, and Alexandre Alahi. Proprio: Latent self- scoring and inference-time refinement for physically plausible video generation, 2026. URL https://arxiv.org/abs/2605.28230

2026 arXiv
[53]

ActiveScope: Actively seeking and correcting perception for MLLMs, 2026

Yajing Wang, Chao Bi, Junshu Sun, Shufan Shen, Zhaobo Qi, Shuhui Wang, and Q Huang. ActiveScope: Actively seeking and correcting perception for MLLMs, 2026. URL https: //arxiv.org/abs/2606.24292

2026 arXiv
[54]

Safe autoregressive image generation with iterative self-improving codebooks, 2026

Yunqi Xue, Zhijiang Li, Philip Torr, and Jindong Gu. Safe autoregressive image generation with iterative self-improving codebooks, 2026. URL https://arxiv.org/abs/2606.27147

2026 arXiv
[55]

Paying more attention to visual tokens in self- evolving large multimodal models, 2026

Shravan Venkatraman, Ritesh Thawkar, Omkar Thawakar, Rao Muhammad Anwer, Hisham Cholakkal, Salman Khan, and Fahad Khan. Paying more attention to visual tokens in self- evolving large multimodal models, 2026. URL https://arxiv.org/abs/2606.27373

2026 arXiv
[56]

Personal visual context learning in large multimodal models, 2026

Zihui Xue, Ami Baid, Sangho Kim, Mi Luo, and Kristen Grauman. Personal visual context learning in large multimodal models, 2026. URL https://arxiv.org/abs/2605.10936

2026 arXiv
[57]

Each judge its own yardstick: Discovering per-VLM taxonomies for physical video evaluation, 2026

Yu Cao, Ziquan Liu, Zhensong Zhang, Jiankang Deng, Shaogang Gong, and Jifei Song. Each judge its own yardstick: Discovering per-VLM taxonomies for physical video evaluation, 2026. URL https://arxiv.org/abs/2606.22918

2026 arXiv
[58]

Query-conditioned test-time self-training for large language models, 2026

Chaehee Song, Minseok Seo, Yeeun Seong, Doyi Kim, and Changick Kim. Query-conditioned test-time self-training for large language models, 2026. URL https://arxiv.org/abs/2605.133 69. 32

2026
[59]

Continual self-improvement with lightweight experiential latent memories, 2026

Vaggelis Dorovatas, Nancy Kalaj, and Rahaf Aljundi. Continual self-improvement with lightweight experiential latent memories, 2026. URL https://arxiv.org/abs/2606.17803

2026 arXiv
[60]

Language models need sleep: Learning to self-modify and consolidate memories, 2026

Ali Behrouz, Farnoosh Hashemi, and Vahab Mirrokni. Language models need sleep: Learning to self-modify and consolidate memories, 2026. URL https://arxiv.org/abs/2606.03979

2026 arXiv
[61]

Truly self-improving agents require intrinsic metacognitive learning, 2025

Tennison Liu and Mihaela van der Schaar. Truly self-improving agents require intrinsic metacognitive learning, 2025. URL https://arxiv.org/abs/2506.05109

2025 arXiv
[62]

Environment-grounded automated prompt optimization for LLM game agents, 2026

Rean Clive Fernandes, Lukas Fehring, Theresa Eimer, Marius Lindauer, and Matthias Feurer. Environment-grounded automated prompt optimization for LLM game agents, 2026. URL https://arxiv.org/abs/2606.17838

2026 arXiv
[63]

Yuxuan Wan, Tianqing Fang, Zaitang Li, Yintong Huo, Wenxuan Wang, Haitao Mi, Dong Yu, and Michael R. Lyu. Inference-time scaling of verification: Self-evolving deep research agents via test-time rubric-guided verification, 2026. URL https://arxiv.org/abs/2601.15808

2026 arXiv
[64]

Shen, Daniel Burkhardt, Meghdad Kurmanji, Nurbek Tastan, Lorenzo Sani, Niccolò Alberto Elia Venanzi, Ambroise Odonnat, Zeyu Cao, Bill Marino, Xinchi Qiu, and Nicholas D

Alex Iacob, Andrej Jovanović, William F. Shen, Daniel Burkhardt, Meghdad Kurmanji, Nurbek Tastan, Lorenzo Sani, Niccolò Alberto Elia Venanzi, Ambroise Odonnat, Zeyu Cao, Bill Marino, Xinchi Qiu, and Nicholas D. Lane. The red queen gödel machine: Co-evolving agents and their ev...

2026 arXiv
[65]

Sc.) Yang

Ziyang Liu, Xinyan Guo, Xuchen Wei, Han Hao, and Liu (M. Sc.) Yang. Escher-loop: Mutual evolution by closed-loop self-referential optimization, 2026. URL https://arxiv.org/abs/2604 .23472

2026
[66]

QueenBee planner: Skill-evolving com- munication topologies for token-eﬀicient LLM multi-agent systems, 2026

Congjia Tian, Yuhang Yao, and Jiaming Cui. QueenBee planner: Skill-evolving com- munication topologies for token-eﬀicient LLM multi-agent systems, 2026. URL https: //arxiv.org/abs/2606.27492

2026 arXiv
[67]

Learn- ing from failure: Inference-time self-improvement for computer-use agents, 2026

Xueqiao Sun, Xiaohan Wang, Ludwig Schmidt, Serena Yeung-Levy, and Yuhui Zhang. Learn- ing from failure: Inference-time self-improvement for computer-use agents, 2026. URL https://arxiv.org/abs/2606.31270

2026 arXiv
[68]

Gang Liao, Y He, Abdullah Ozturk, Zhouyang Li, Ying Wang, Zichang Guo, Hongsen Qin, Yaobin Qin, Tao Yang, Zewei Jiang, Dianshi Li, Jort Gemmeke, Jiangyuan Li, Liyuan Li, Nathan Yan, Masha Basmanova, Uladzimir Pashkevich, Matt Steiner, Pedro Pedreira, Rob Fergus, Anirudh Goyal,...

2026 arXiv
[69]

The meta-agent challenge: Are current agents capable of autonomous agent development?, 2026

Xinyu Lu, Tianshu Wang, Pengbo Wang, Zujie Wen, Zhiqiang Zhang, Jun Zhou, Boxi Cao, Yaojie Lu, Hongyu Lin, Xianpei Han, and Le Sun. The meta-agent challenge: Are current agents capable of autonomous agent development?, 2026. URL https://arxiv.org/abs/2606.0 4455

2026
[70]

SAGE: A quantitative evaluation of socialized evolution in agent ecosystems, 2026

Linyue Pan, Yaoming Zhu, Lin Qiu, Xuezhi Cao, and Xunliang Cai. SAGE: A quantitative evaluation of socialized evolution in agent ecosystems, 2026. URL https://arxiv.org/abs/26 06.03544

2026
[71]

Multi-agent reasoning improves compute eﬀiciency: Pareto-optimal test-time scaling, 2026

Florian Valentin Wunderlich, Lars Benedikt Kaesberg, Jan Philip Wahle, Terry Ruas, and Bela Gipp. Multi-agent reasoning improves compute eﬀiciency: Pareto-optimal test-time scaling, 2026. URL https://arxiv.org/abs/2605.01566. 33

2026 arXiv
[72]

Voyager: An open-ended embodied agent with large language models, 2023

Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi Fan, and Anima Anandkumar. Voyager: An open-ended embodied agent with large language models, 2023. URL https://arxiv.org/abs/2305.16291. TMLR 2024

2023 arXiv
[73]

SkillAxe: Sharpening LLM- authored agent skills through evaluation-guided self-refinement, 2026

Srishti Gautam, Arjun Radhakrishna, and Sumit Gulwani. SkillAxe: Sharpening LLM- authored agent skills through evaluation-guided self-refinement, 2026. URL https://arxi v.org/abs/2606.10546

2026 arXiv
[74]

Rossi, Jingbo Shang, Julian McAuley, and Junda Wu

Yash Vishe, Rohan Surana, Xunyi Jiang, Zihan Huang, Xintong Li, Nikki Lijing Kuang, Tong Yu, Ryan A. Rossi, Jingbo Shang, Julian McAuley, and Junda Wu. Skill-r1: Agent skill evolution via reinforcement learning, 2026. URL https://arxiv.org/abs/2605.09359

2026 arXiv
[75]

SkillRevise: Improving LLM-authored agent skills via trace-conditioned skill revision, 2026

Yuxuan Liu, Zhaochen Su, Lingyun Xie, Yuhao Zhang, Qing Zong, Jiahe Guo, Zhongwei Xie, Yiyan Ji, Yauwai Yim, Hongyu Luo, Xiyu Ren, Ruan Chenyu, Haoran Li, and Yangqiu Song. SkillRevise: Improving LLM-authored agent skills via trace-conditioned skill revision, 2026. URL https:/...

2026 arXiv
[76]

AlgoSkill: Learning to design algorithms by scheduling human-like skills, 2026

Xinyuan Song, Zekun Cai, and L Zhao. AlgoSkill: Learning to design algorithms by scheduling human-like skills, 2026. URL https://arxiv.org/abs/2606.29999

2026 arXiv
[77]

SkillMaster: Toward autonomous skill mastery in LLM agents, 2026

Min Yang, Jinghua Piao, Xu Xia, Xiaochong Lan, Jiaju Chen, Yongshun Gong, and Yong Li. SkillMaster: Toward autonomous skill mastery in LLM agents, 2026. URL https://arxiv.or g/abs/2605.08693

2026 arXiv
[78]

FederatedSkill: Federated learning for agentic skill evolution, 2026

Jingbo Yang, Guanyu Yao, Yang Zhang, Ramana Rao Kompella, Gaowen Liu, and Shiyu Chang. FederatedSkill: Federated learning for agentic skill evolution, 2026. URL https: //arxiv.org/abs/2606.03143

2026 arXiv
[79]

SkillSmith: Co-evolving skills and tools for self-improving agent systems, 2026

Yangbo Wei, Zhen Huang, Shaoqiang Lu, Junhong Qian, Qifan Wang, Chen Wu, and Lei He. SkillSmith: Co-evolving skills and tools for self-improving agent systems, 2026. URL https://arxiv.org/abs/2606.01314

2026 arXiv
[80]

Socratic-SWE: Self-evolving coding agents via trace-derived agent skills, 2026

Chuan Xiao, Zhengbo Jiao, Shaobo Wang, Wei Wang, Bing Zhao, Hu Wei, Linfeng Zhang, and Lin Qu. Socratic-SWE: Self-evolving coding agents via trace-derived agent skills, 2026. URL https://arxiv.org/abs/2606.07412

2026 arXiv
[81]

SHARP: A self-evolving human-auditable rubric policy for financial trading agents, 2026

Xiwen Chen, Wenhui Zhu, Songzhu Zheng, Kashif Rasul, Yueyue Deng, and Huayu Li. SHARP: A self-evolving human-auditable rubric policy for financial trading agents, 2026. URL https://arxiv.org/abs/2605.06822

2026 arXiv
[82]

SkillMutator: Benchmarking and defending language-and-code cross-modal attacks on LLM agent skills, 2026

Youngduk Kim, Minkyoo Song, and Seungwon Shin. SkillMutator: Benchmarking and defending language-and-code cross-modal attacks on LLM agent skills, 2026. URL https: //arxiv.org/abs/2606.14154

2026
[83]

SkillHarness: Harnessing safe skills for computer-use agents, 2026

Yurun Chen, Biao Yi, Keting Yin, and Shengyu Zhang. SkillHarness: Harnessing safe skills for computer-use agents, 2026. URL https://arxiv.org/abs/2606.20636

2026 arXiv
[84]

Bhatt, Kevin Wang, Samuel Tetteh, Zhangyang Wang, and Ufuk Topcu

Yunhao Yang, Neel P. Bhatt, Kevin Wang, Samuel Tetteh, Zhangyang Wang, and Ufuk Topcu. V ASO: Formally verifiable self-evolving skills for physical AI agents, 2026. URL https://arxiv.org/abs/2606.05395

2026 arXiv
[85]

Safety in self-evolving LLM agent systems: Threats, amplification, and case studies, 2026

Ruixiao Lin, Xinhao Deng, Qingming Li, Jianan Ma, Yunhao Feng, Yuqi Qing, Zhenyuan Li, Yechao Zhang, Shiwen Cui, Changhua Meng, Tianwei Zhang, Xingjun Ma, Qi Li, Ke Xu, 34 and Shouling Ji. Safety in self-evolving LLM agent systems: Threats, amplification, and case studies, 202...

2026 arXiv
[86]

Towards healthy evolution: Exploring the role and mechanisms of human-agent interaction in self-evolving systems, 2026

Dongyong Shi, Junqi He, Junhao Chen, Bowen Wang, and Yuta Nakashima. Towards healthy evolution: Exploring the role and mechanisms of human-agent interaction in self-evolving systems, 2026. URL https://arxiv.org/abs/2606.06114

2026 arXiv
[87]

Co-Reyes, Rishabh Agarwal, et al

A vi Singh, John D. Co-Reyes, Rishabh Agarwal, et al. Beyond human data: Scaling self- training for problem-solving with language models, 2023. URL https://arxiv.org/abs/2312.0

2023
[88]

Self-play fine-tuning converts weak language models to strong language models

Zixiang Chen, Yihe Deng, Huizhuo Yuan, Kaixuan Ji, and Quanquan Gu. Self-play fine-tuning converts weak language models to strong language models. In International Conference on Machine Learning (ICML) , 2024. URL https://arxiv.org/abs/2401.01335. SPIN

2024 arXiv
[89]

ReST- MCTS*: LLM self-training via process reward guided tree search, 2024

Dan Zhang, Sining Zhoubian, Ziniu Hu, Yisong Yue, Yuxiao Dong, and Jie Tang. ReST- MCTS*: LLM self-training via process reward guided tree search, 2024. URL https://arxiv. org/abs/2406.03816

2024 arXiv
[90]

SEV A: Self-evolving verification agent with process reward for fact attribution, 2026

Aojie Yuan, Yi Nian, Haiyue Zhang, Zijian Su, and Yue Zhao. SEV A: Self-evolving verification agent with process reward for fact attribution, 2026. URL https://arxiv.org/abs/2606.29713

2026 arXiv
[91]

EvoIdeator: Evolving scientific ideas through checklist-grounded reinforcement learning, 2026

Andreas Sauter, Yuyue Zhao, Jacopo Urbani, Wenxiang Hu, Zaiqiao Meng, Lun Zhou, Xiao- hui Yan, and Yougang Lyu. EvoIdeator: Evolving scientific ideas through checklist-grounded reinforcement learning, 2026. URL https://arxiv.org/abs/2603.21728

2026
[93]

The value axis: Language models encode whether they’re on the right track, 2026

Nick Jiang, Isaac Kauvar, and Jack Lindsey. The value axis: Language models encode whether they’re on the right track, 2026. URL https://arxiv.org/abs/2606.17056

2026
[94]

Self-trained verification for training- and test-time self-improvement, 2026

Chen Henry Wu and Aditi Raghunathan. Self-trained verification for training- and test-time self-improvement, 2026. URL https://arxiv.org/abs/2605.30290

2026 arXiv
[95]

Self-improvement can self-regress: The rise-and-collapse failure mode of LLM self-training, 2026

Jianzhe Lin. Self-improvement can self-regress: The rise-and-collapse failure mode of LLM self-training, 2026. URL https://arxiv.org/abs/2606.21090

2026 arXiv
[96]

When LLM reward design fails: Diagnostic-driven refinement for sparse structured RL, 2026

Youting Wang, Yuan Tang, Bowen Liu, Xuan Liu, and Dingyan Shang. When LLM reward design fails: Diagnostic-driven refinement for sparse structured RL, 2026. URL https://arxi v.org/abs/2605.28918

2026 arXiv
[97]

Repeated post-training is not self-improving: Diagnosing scientific amnesia in continual DPO pipelines,

Jianzhe Lin, Fang Wang, Xiaolin Li, Rajeshkumar Golani, and Jubin Chheda. Repeated post-training is not self-improving: Diagnosing scientific amnesia in continual DPO pipelines,
[98]

URL https://arxiv.org/abs/2606.21089

arXiv
[99]

Re-ReST: Reflection-reinforced self-training for language agents, 2024

Zi-Yi Dou, Cheng-Fu Yang, Xueqing Wu, Kai-Wei Chang, and Nanyun Peng. Re-ReST: Reflection-reinforced self-training for language agents, 2024. URL https://arxiv.org/abs/24 06.01495

2024
[100]

Markus J. Buehler. PRefLexOR: Preference-based recursive language modeling for ex- ploratory optimization of reasoning and agentic thinking, 2024. URL https://arxiv.org/ abs/2410.12375. 35

2024 arXiv
[101]

Language models are hidden reasoners: Unlocking latent reasoning capabilities via self-rewarding, 2024

Haolin Chen, Yihao Feng, Zuxin Liu, Weiran Yao, Akshara Prabhakar, Shelby Heinecke, Ricky Ho, Phil Mui, Silvio Savarese, Caiming Xiong, and Huan Wang. Language models are hidden reasoners: Unlocking latent reasoning capabilities via self-rewarding, 2024. URL https://arxiv.org/...

2024 arXiv
[102]

NeuReasoner: Theory-grounded mapping of reasoning elicitation bound- aries, 2026

Aydin Javadov, Shyngys Aitkazinov, Tobias Hoesli, Florian von Wangenheim, Bjoern Schuller, and James Ollier. NeuReasoner: Theory-grounded mapping of reasoning elicitation bound- aries, 2026. URL https://arxiv.org/abs/2606.29971

2026 arXiv
[103]

Self-verified distillation: Your language model is secretly its own synthetic data pipeline, 2026

Tony Lee and Percy Liang. Self-verified distillation: Your language model is secretly its own synthetic data pipeline, 2026. URL https://arxiv.org/abs/2605.26132

2026 arXiv
[104]

He, Jiuming Liu, and Victor Sanchez

H. He, Jiuming Liu, and Victor Sanchez. Revisiting chain-of-thought reasoning under limited supervision: Semi-supervised chain-of-thought learning, 2026. URL https://arxiv.org/abs/ 2607.01511

2026 arXiv
[105]

Do NOT think that much for 2+3=? on the overthinking of o1-like LLMs, 2024

Xingyu Chen, Jiahao Xu, Tian Liang, Zhiwei He, Ji Pang, Dian Yu, Linfeng Song, Qiuzhi Liu, Mengfei Zhou, Zhuosheng Zhang, Rui Wang, Zhaopeng Tu, Haitao Mi, and Dongbo Yu. Do NOT think that much for 2+3=? on the overthinking of o1-like LLMs, 2024. URL https://arxiv.org/abs/2412.21187

2024 arXiv
[106]

Know when to stop: Segment-level credit assignment for reducing overthinking, 2026

C Chung min Lee, Sihui Dai, Mingyang Zhou, Isha Slavin, Shi-Xiong Zhang, Sambit Sahu, and William Campbell. Know when to stop: Segment-level credit assignment for reducing overthinking, 2026. URL https://arxiv.org/abs/2607.00482

2026 arXiv
[107]

Difan Jiao, Q. G. Wen, Blair Yang, Zhenwei Tang, and Ashton Anderson. ThinkTwice: Jointly optimizing large language models for reasoning and self-refinement, 2026. URL https: //arxiv.org/abs/2604.01591

2026 arXiv
[108]

A verifiable search is not a learnable chain-of-thought, 2026

Harsh Patel. A verifiable search is not a learnable chain-of-thought, 2026. URL https: //arxiv.org/abs/2606.21884

2026 arXiv
[109]

Power distribution bridges sampling, self-reward RL, and self-distillation, 2026

Akiyoshi Tomihari and Issei Sato. Power distribution bridges sampling, self-reward RL, and self-distillation, 2026. URL https://arxiv.org/abs/2605.04542

2026 arXiv
[110]

Embarrassingly simple self-distillation improves code generation, 2026

Ruixiang Zhang, Richard He Bai, Huangjie Zheng, Navdeep Jaitly, Ronan Collobert, and Yizhe Zhang. Embarrassingly simple self-distillation improves code generation, 2026. URL https://arxiv.org/abs/2604.01193

2026 arXiv
[111]

ReNIO: Reweighting negative trajectory importance for LLM on-policy distillation, 2026

C Y Lin, Kedi Chen, and Wei Zhang. ReNIO: Reweighting negative trajectory importance for LLM on-policy distillation, 2026. URL https://arxiv.org/abs/2606.23104

2026 arXiv
[112]

DRIFT: Diﬀiculty routing self-DIstillation with rhythm-gated exploration and success BuFfer training, 2026

Haisen Luo, Yiwei Liu, Haoning Wang, Dan Liu, Junxi Yin, Haotian Wang, Lei Zhang, Xiaoyu Tian, Shuaiting Chen, Yuansheng Song, Baoyan Guo, Xiongfei Yan, B Yang, Chengwei Liu, Ming Cui, and Jiong Chen. DRIFT: Diﬀiculty routing self-DIstillation with rhythm-gated exploration and...

2026 arXiv
[113]

Procedural memory distillation: Online reflection for self-improving language models, 2026

Ye Liu, Srijan Bansal, Bo Pang, Yang Li, Zeyu Leo Liu, Yifei Ming, Zixuan Ke, Shafiq Joty, and Semih Yavuz. Procedural memory distillation: Online reflection for self-improving language models, 2026. URL https://arxiv.org/abs/2607.01480

2026 arXiv
[114]

Purified OPSD: On-policy self-distillation without losing how to think, 2026

Zhanming Shen, Jintao Tong, Su Yan, Chen Shen, Hao Chen, Wentao Ye, Xiaomeng Hu, Rui Miao, Haobo Wang, Junbo Zhao, Gang Chen, and J P Ye. Purified OPSD: On-policy self-distillation without losing how to think, 2026. URL https://arxiv.org/abs/2607.02234. 36

2026 arXiv
[115]

DemoPSD: Disagreement-modulated policy self-distillation, 2026

Yunhe Li, Hao Shi, Wenhao Liu, Mengzhe Ruan, Hanxu Hou, Zhongxiang Dai, Shuang Qiu, and Linqi Song. DemoPSD: Disagreement-modulated policy self-distillation, 2026. URL https://arxiv.org/abs/2607.02502

2026 arXiv
[116]

RLCSD: Reinforcement learning with contrastive on-policy self-distillation,

Leyi Pan, Shuchang Tao, Yunpeng Zhai, Lingzhe Zhang, Zhaoyang Liu, Bolin Ding, Aiwei Liu, and Lijie Wen. RLCSD: Reinforcement learning with contrastive on-policy self-distillation,
[117]

URL https://arxiv.org/abs/2606.11709

arXiv
[118]

Trajectory-refined distillation, 2026

Li Jiang, Haoran Xu, Yichuan Ding, and Amy Zhang. Trajectory-refined distillation, 2026. URL https://arxiv.org/abs/2606.08432

2026 arXiv
[119]

The role of feedback alignment in self-distillation, 2026

Semih Kara and Oğuzhan Ersoy. The role of feedback alignment in self-distillation, 2026. URL https://arxiv.org/abs/2606.11173

2026 arXiv
[120]

Seeing before reasoning: Decoupling perception and reasoning for shortcut-resilient multimodal on-policy self-distillation, 2026

Sihan Wang, Xiyao Liu, Lianqing Liu, and Zhi Han. Seeing before reasoning: Decoupling perception and reasoning for shortcut-resilient multimodal on-policy self-distillation, 2026. URL https://arxiv.org/abs/2606.19120

2026 arXiv
[121]

Agent0: Unleashing self-evolving agents from zero data via tool-integrated reasoning, 2025

Peng Xia, Kaide Zeng, Jiaqi Liu, Can Qin, Fang Wu, Yiyang Zhou, Caiming Xiong, and Huaxiu Yao. Agent0: Unleashing self-evolving agents from zero data via tool-integrated reasoning, 2025. URL https://arxiv.org/abs/2511.16043. ICML 2026

2025
[122]

Gummadi, and Meeyoung Cha

Kyeongjin Ahn, Seungeon Lee, Krishna P. Gummadi, and Meeyoung Cha. GeoX: Mastering geospatial reasoning through self-play and verifiable rewards, 2026. URL https://arxiv.org/ abs/2605.20006

2026 arXiv
[123]

A theoretical framework for self-play theorem proving algo- rithms, 2026

Thomas Chen and Zhiyuan Li. A theoretical framework for self-play theorem proving algo- rithms, 2026. URL https://arxiv.org/abs/2606.01861

2026 arXiv
[124]

Verifier-backed hard problem generation for mathematical reasoning, 2026

Yuhang Lai, Jiazhan Feng, Yee Whye Teh, and Ning Miao. Verifier-backed hard problem generation for mathematical reasoning, 2026. URL https://arxiv.org/abs/2605.06660

2026 arXiv
[125]

Sophia Xiao Pu, Zhaotian Weng, Chengzhi Liu, Jayanth Srinivasa, Gaowen Liu, William W. B. Wang, and Xin Eric Wang. Survive or collapse: The asymmetric roles of data gating and reward grounding in self-play RL, 2026. URL https://arxiv.org/abs/2605.22217

2026 arXiv
[126]

Mavor- Parker, and Matthew James Sargent

Roger Creus Castanyer, Geoffrey Bradway, Lorenz Wolf, Maxwill Lin, Augustine N. Mavor- Parker, and Matthew James Sargent. PopuLoRA: Co-evolving LLM populations for reasoning self-play, 2026. URL https://arxiv.org/abs/2605.16727

2026 arXiv
[127]

Conversation for non-verifiable learning: Self-evolving LLMs through meta-evaluation

Yuan Sui and Bryan Hooi. Conversation for non-verifiable learning: Self-evolving LLMs through meta-evaluation. Open MIND , 2026. doi: 10.48550/arxiv.2601.21464. URL https: //arxiv.org/abs/2601.21464

2026 doi arXiv
[128]

G-zero: Self-play for open-ended generation from zero data, 2026

Chengsong Huang, Haolin Liu, Tong Zheng, Runpeng Dai, Langlin Huang, Jinyuan Li, Zongxia Li, Zhepei Wei, Yu Meng, and Jiaxin Huang. G-zero: Self-play for open-ended generation from zero data, 2026. URL https://arxiv.org/abs/2605.09959

2026 arXiv
[129]

Ko and Jonas Geiping

T.P. Ko and Jonas Geiping. Attractor states emerge in multi-turn LLM conversations, 2026. URL https://arxiv.org/abs/2606.30571

2026 arXiv
[130]

The robot’s inner critic: Self- refinement of social behaviors through VLM-based replanning, 2026

Jiyu Kelly Lim, Youngwoo Yoon, and Kwanghyun Park. The robot’s inner critic: Self- refinement of social behaviors through VLM-based replanning, 2026. URL https://arxiv.or g/abs/2603.20164. 37

2026
[131]

PhysReflect-VLA: Physical feasibility and self-reflective regulation for reliable vision- language-action policies, 2026

Jiayu Yang, Tao Yang, Weijun Li, Xiang Chang, Fei Chao, Changjing Shang, and Qiang Shen. PhysReflect-VLA: Physical feasibility and self-reflective regulation for reliable vision- language-action policies, 2026. URL https://arxiv.org/abs/2606.27146

2026 arXiv
[132]

Embodied-r1.5: Evolving physical intelligence via embodied foundation models,

Yifu Yuan, Yaoting Huang, Xianze Yao, Yutong Li, Shuoheng Zhang, Linqi Han, Pengyi Li, Jiangeng Sun, Wenting Jia, Zhao Zhang, Yuhao Liu, Ruihao Liao, Yucheng Hu, Qiyu Wu, Yuxiao Li, Zibin Dong, Fei Ni, Yan Zheng, Shuyang Gu, Yi Ma, Hongyao Tang, Han Hu, and Jianye Hao. Embodie...
[133]

URL https://arxiv.org/abs/2606.11324

arXiv
[134]

Robot self-improvement via human-video dynamics models,

Hanzhi Chen, A M Zhang, Simon Schaefer, Kejia Chen, Shi Chen, Daniel Cremers, Oier Mees, and Stefan Leutenegger. Robot self-improvement via human-video dynamics models,
[135]

URL https://arxiv.org/abs/2606.21406

arXiv
[136]

DataEvolver: Automatic data preparation for large language models through multi-level self-evolving, 2026

Chao Deng, Shaolei Zhang, Ju Fan, and Xiaoyong Du. DataEvolver: Automatic data preparation for large language models through multi-level self-evolving, 2026. URL https: //arxiv.org/abs/2606.07001

2026 arXiv
[137]

Let’s verify step by step, 2023

Hunter Lightman, Vineet Kosaraju, Yura Burda, Harri Edwards, Bowen Baker, Teddy Lee, Jan Leike, John Schulman, Ilya Sutskever, and Karl Cobbe. Let’s verify step by step, 2023. URL https://arxiv.org/abs/2305.20050. ICLR 2024

2023 arXiv
[138]

Xing, Hao Zhang, Joseph E

Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. Judging LLM-as-a-judge with MT-bench and chatbot arena. In Advances in Neural Information Process...

2023 arXiv
[139]

Scaling laws for reward model overoptimization

Leo Gao, John Schulman, and Jacob Hilton. Scaling laws for reward model overoptimization. In International Conference on Machine Learning (ICML) , 2023. URL https://arxiv.org/ab s/2210.10760

2023 arXiv
[141]

LLM judges have dark current: A psychometric datasheet for LLM-as-a-judge evaluation, 2026

Hiroyasu Usami, Keisuke Hara, Ayato Tsuboi, and Naohiko Matsuda. LLM judges have dark current: A psychometric datasheet for LLM-as-a-judge evaluation, 2026. URL https: //arxiv.org/abs/2606.15610

2026
[142]

When languages disagree: Self-evolving multilingual LLM judges,

Xiyan Fu and Wei Lu. When languages disagree: Self-evolving multilingual LLM judges,
[143]

URL https://arxiv.org/abs/2606.08092

arXiv
[144]

Support vector rubrics: Closing the gap between self-generated and human rubrics, 2026

Mengyuan Sun, Yu Li, Zhuohao Yu, Shikun Zhang, and Wei Ye. Support vector rubrics: Closing the gap between self-generated and human rubrics, 2026. URL https://arxiv.org/ab s/2606.08077

2026 arXiv
[146]

Rethinking reward super- vision: Rubric-conditioned self-distillation, 2026

Siyi Gu, Jialin Chen, Sophia Zhou, Arman Cohan, and Rex Ying. Rethinking reward super- vision: Rubric-conditioned self-distillation, 2026. URL https://arxiv.org/abs/2606.19327

2026 arXiv
[147]

Lu, and Yuxuan Wang

Shuaijie She, Yu Bao, Lu Yu, Lu Xu, Tao Li, Wenhao Zhu, Shujian Huang, Shanbo Cheng, L. Lu, and Yuxuan Wang. DuPO: Enabling reliable LLM self-verification via dual preference optimization, 2025. URL https://arxiv.org/abs/2508.14460

2025 arXiv
[148]

The mirror loop: Recursive non-convergence in generative reasoning sys- tems, 2025

Bentley DeVilling. The mirror loop: Recursive non-convergence in generative reasoning sys- tems, 2025. URL https://arxiv.org/abs/2510.21861

2025
[149]

Breaking the self-confirming loop: Diagnosing and mitigating systemic reward bias in self-rewarding RL, 2025

Chuyi Tan, Peiwen Yuan, Xinglin Wang, Yicheng Li, Shaoxiong Feng, Yueqi Zhang, Jiayi Shi, Ji Zhang, Boyuan Pan, Yao Hu, and Kan Li. Breaking the self-confirming loop: Diagnosing and mitigating systemic reward bias in self-rewarding RL, 2025. URL https://arxiv.org/ab s/2510.08977

2025 arXiv
[150]

Deterministic integrity gates for LLM-assisted clinical manuscript preparation: An auditable biomedical informatics architecture, 2026

Yoojin Nam, Jinhoon Jeong, and Namkug Kim. Deterministic integrity gates for LLM-assisted clinical manuscript preparation: An auditable biomedical informatics architecture, 2026. URL https://arxiv.org/abs/2606.09500

2026 arXiv
[151]

SciIntegrity-bench: A benchmark for evaluating academic integrity in AI scientist systems, 2026

Zonglin Yang, Xingtong Liu, and Xinyan Xu. SciIntegrity-bench: A benchmark for evaluating academic integrity in AI scientist systems, 2026. URL https://arxiv.org/abs/2605.10246

2026 arXiv
[152]

AI models collapse when trained on recursively generated data

Ilia Shumailov, Zakhar Shumaylov, Yiren Zhao, Nicolas Papernot, Ross Anderson, and Yarin Gal. AI models collapse when trained on recursively generated data. Nature, 631:755–759,
[153]

doi: 10.1038/s41586-024-07566-y

doi
[154]

Entropy-reservoir bregman projection: An information-geometric unification of model collapse, 2025

Jingwei Chen. Entropy-reservoir bregman projection: An information-geometric unification of model collapse, 2025. URL https://arxiv.org/abs/2512.14879

2025
[155]

On the limits of self-improving in large language models: The singularity is not near without symbolic model synthesis, 2026

Hector Zenil. On the limits of self-improving in large language models: The singularity is not near without symbolic model synthesis, 2026. URL https://arxiv.org/abs/2601.05280

2026
[156]

Generating on generated: An approach towards self-evolving diffusion models, 2025

Xulu Zhang, Xiaoyong Wei, Jinlin Wu, Jiaxin Wu, Zhaoxiang Zhang, Zhen Lei, and Qing Li. Generating on generated: An approach towards self-evolving diffusion models, 2025. URL https://arxiv.org/abs/2502.09963

2025 arXiv
[157]

Better, faster: Harnessing self-improvement in large reasoning models, 2026

Qihuang Zhong, Liang Ding, Juhua Liu, Bo Du, Leszek Rutkowski, and Dacheng Tao. Better, faster: Harnessing self-improvement in large reasoning models, 2026. URL https://arxiv.or g/abs/2605.24998

2026 arXiv
[158]

Vocabulary dropout for curriculum diversity in LLM co-evolution, 2026

Jacob Dineen, Aswin RR V, Zhikun Xu, and Ben Zhou. Vocabulary dropout for curriculum diversity in LLM co-evolution, 2026. URL https://arxiv.org/abs/2604.03472

2026 arXiv
[159]

From fluent to verifiable: Claim-level auditability for deep research agents

Razeen A Rasheed, Somnath Banerjee, Animesh Mukherjee, and Rima Hazra. From fluent to verifiable: Claim-level auditability for deep research agents. Open MIND , 2026. doi: 10.48550/arxiv.2602.13855. URL https://arxiv.org/abs/2602.13855

2026 doi
[160]

ISM:self-improving strategy memory for continual mathemat- ical reasoning, 2026

Prakhar Dixit and Tim Oates. ISM:self-improving strategy memory for continual mathemat- ical reasoning, 2026. URL https://arxiv.org/abs/2606.31191

2026 arXiv
[161]

ExpGraph: Model-agnostic experience learning with graph-structured memory for LLM agents, 2026

Tao Feng, Chongrui Ye, Tianyang Luo, Jingjun Xu, Xueqiang Xu, Haozhen Zhang, Zhigang Hua, Yan Xie, Shuang Yang, Ge Liu, and Jiaxuan You. ExpGraph: Model-agnostic experience learning with graph-structured memory for LLM agents, 2026. URL https://arxiv.org/abs/ 2605.30712. 39

2026 arXiv
[162]

Knowledge-graph paths as intermediate supervision for self-evolving search agents, 2026

Huyu Wu, Jun Liu, Xiaochi Wei, Yan Gao, Yi Wu, and Yao Hu. Knowledge-graph paths as intermediate supervision for self-evolving search agents, 2026. URL https://arxiv.org/abs/ 2605.05702

2026 arXiv
[163]

Evolution of heuristics: Towards eﬀicient automatic algorithm design using large language model, 2024

Fei Liu, Xialiang Tong, Mingxuan Yuan, Xi Lin, Fu Luo, Zhenkun Wang, Zhichao Lu, and Qingfu Zhang. Evolution of heuristics: Towards eﬀicient automatic algorithm design using large language model, 2024. URL https://arxiv.org/abs/2401.02051

2024 arXiv
[164]

QUBE: Enhancing automatic heuristic design via quality-uncertainty balanced evolution, 2024

Zijie Chen, Zhanchao Zhou, Yu Lü, Renjun Xu, Lili Pan, and Zhenzhong Lan. QUBE: Enhancing automatic heuristic design via quality-uncertainty balanced evolution, 2024. URL https://arxiv.org/abs/2412.20694

2024 arXiv
[165]

Doan, and Huynh Thi Thanh Binh

Pham Vu Tuan Dat, L. Doan, and Huynh Thi Thanh Binh. HSEvo: Elevating automatic heuristic design with diversity-driven harmony search and genetic algorithm using LLMs,
[166]

URL https://arxiv.org/abs/2412.14995

arXiv
[167]

DEI: Diversity in evolutionary inference for quality- diversity search, 2026

John Donaghy and Shikhar Rastogi. DEI: Diversity in evolutionary inference for quality- diversity search, 2026. URL https://arxiv.org/abs/2605.27130

2026 arXiv
[168]

AI-PROPELLER: Warehouse-scale inter- procedural code layout optimization with AlphaEvolve, 2026

Chaitanya Mamatha Ananda, Rajiv Gupta, Mircea Trofin, Aiden Grossman, Sriraman Tal- lam, Xinliang David Li, and Amir Yazdanbakhsh. AI-PROPELLER: Warehouse-scale inter- procedural code layout optimization with AlphaEvolve, 2026. URL https://arxiv.org/abs/26 06.00131

2026
[169]

Adapt- ing AlphaEvolve to optimize fully homomorphic encryption on TPUs, 2026

Shruthi Gorantala, Jianming Tong, Asra Ali, Baiyu Li, Jonathan Katz, Jeremy Kun, Thomas Steinke, Abhradeep Thakurta, Julian Walker, and Amir Yazdanbakhsh. Adapt- ing AlphaEvolve to optimize fully homomorphic encryption on TPUs, 2026. URL https: //arxiv.org/abs/2605.14718

2026 arXiv
[170]

EVOM: Agentic meta-evolution of actor-critic architectures for reinforcement learning, 2026

Boyun Zhang, Chao Wang, and Kai Wu. EVOM: Agentic meta-evolution of actor-critic architectures for reinforcement learning, 2026. URL https://arxiv.org/abs/2606.26327

2026 arXiv
[171]

From AI assistant to AI scientist: Autonomous discovery of LLM-RL algorithms with LLM agents, 2026

Sirui Xia, Yikai Zhang, Aili Chen, Siye Wu, Siyu Yuan, and Yanghua Xiao. From AI assistant to AI scientist: Autonomous discovery of LLM-RL algorithms with LLM agents, 2026. URL https://arxiv.org/abs/2603.23951

2026
[172]

MLEvolve: A self-evolving framework for automated machine learning algorithm discovery, 2026

Shangheng Du, Xiangchao Yan, Jinxin Shi, Zongsheng Cao, Shiyang Feng, Zichen Liang, Boyuan Sun, Tianshuo Peng, Yifan Zhou, Xin Li (51274), Jie Zhou, Liang He, Bo Zhang, and Lei Bai. MLEvolve: A self-evolving framework for automated machine learning algorithm discovery, 2026. U...

2026 arXiv
[173]

Scientific discovery as meta- optimization: a combinatorial optimization case study, 2026

Yuanhang Zhang, Chesson Sipling, and Massimiliano Di Ventra. Scientific discovery as meta- optimization: a combinatorial optimization case study, 2026. URL https://arxiv.org/abs/26 06.26728

2026
[174]

Li, Wenjie Ma, Li Cao, Ziheng Zhou, Mert Cemri, Shu Liu, Yuran Xiu, Chenxiao Yan, Haikun Zhao, Bin Yu, Ion Stoica, and Dawn Song

Yangzhen Wu, Aaron J. Li, Wenjie Ma, Li Cao, Ziheng Zhou, Mert Cemri, Shu Liu, Yuran Xiu, Chenxiao Yan, Haikun Zhao, Bin Yu, Ion Stoica, and Dawn Song. BenchEvolver: Frontier task synthesis via solution-centric evolution, 2026. URL https://arxiv.org/abs/2606.01286

2026 arXiv
[176]

LEVI: Stronger search architectures can substitute for larger LLMs in evolutionary search, 2026

Temoor Tanveer. LEVI: Stronger search architectures can substitute for larger LLMs in evolutionary search, 2026. URL https://arxiv.org/abs/2605.09764. 40

2026 arXiv
[177]

Tony Feng, Trieu Trinh, Garrett Bingham, Dawsen Hwang, Yuri Chervonyi, Junehyuk Jung, Joonkyung Lee, Carlo Pagano, Sang hyun Kim, Federico Pasqualotto, Sergei Gukov, Jonathan N. Lee, Junsu Kim, Kaiying Hou, Golnaz Ghiasi, Yi Tay, Yaguang Li, Chenkai Kuang, Yuan Liu, Hanzhao, L...

2026 doi
[178]

Socratic agents for autonomous scientific discovery in high-dimensional physical systems,

Xianrui Zeng, Pengfei Liu, Yirui Zang, Yang Shen, Fei Yu, C Shijia Yu, M Y Liu, and Yang Du. Socratic agents for autonomous scientific discovery in high-dimensional physical systems,
[179]

URL https://arxiv.org/abs/2606.26722

arXiv
[180]

Pi- otrowski, Jakub Kaliski, Jacopo Urbani, Zaiqiao Meng, Lun Zhou, and Xiaohui Yan

Yougang Lyu, Xi Zhang, Xinhao Yi, Yuyue Zhao, Shuyu Guo, Wenxiang Hu, Jan A. Pi- otrowski, Jakub Kaliski, Jacopo Urbani, Zaiqiao Meng, Lun Zhou, and Xiaohui Yan. EvoSci- entist: Towards multi-agent evolving AI scientists for end-to-end scientific discovery. Open MIND, 2026. do...

2026 doi
[181]

EvoMaster: A foundational evolving agent framework for agentic science at scale, 2026

Xinyu Zhu, Yuzhu Cai, Zexi Liu, Cheng Wang, Fengyang Li, Wenkai Jin, Wanxu Liu, Zehao Bing, Bingyang Zheng, Jingyi Chai, Shuo Tang, Rui Ye, Yuwen Du, Xianghe Pang, Yaxin Du, Tingjia Miao, Yuzhi Zhang, Ruoxue Liao, Zhaohan Ding, Linfeng Zhang, Yanfeng Wang, Weinan E, and Siheng...

2026 arXiv
[182]

Clarus: Coordinating autonomous research agents toward web-scale scientific collaboration, 2026

Zihan Guo, Zeyi Chen, Zhiyu Chen, Zicai Cui, Shuai Shao, Bo Huang, Zhi Han, Y Song, Yuan Yuan, Chenxi Zeng, Xiaohang Nie, Zhengxi Yu, Hanwen Zhu, J G Liao, Ming Zhou, Yang Li, Yuanjian Zhou, and Weinan Zhang. Clarus: Coordinating autonomous research agents toward web-scale sci...

2026 arXiv
[183]

EurekAgent: Agent environment engineering is all you need for autonomous scientific discovery, 2026

Amy Xin, Jiening Siow, Junjie Wang, Zijun Yao, F Zhang, Jian Song, Lei Hou, and Juanzi Li. EurekAgent: Agent environment engineering is all you need for autonomous scientific discovery, 2026. URL https://arxiv.org/abs/2606.13662

2026 arXiv
[184]

Heuresis: Search strategies for autonomous AI research agents across quality, diversity and novelty, 2026

Antonis Antoniades, Deepak Nathani, Ritam Saha, Alfonso Amayuelas, Ivan Bercovich, Zhao- tian Weng, Vignesh Baskaran, Kunal Bhatia, and William Yang Wang. Heuresis: Search strategies for autonomous AI research agents across quality, diversity and novelty, 2026. URL https://arx...

2026 arXiv
[185]

Externalizing research synthesis and validation in AI scientists through a research harness, 2026

Zijian Wang, Hanqi Li, Ziyue Yang, Zijian Hu, Shenghan Zuo, Yunzhe Zhang, Da Ma, Danyu Luo, Chenrun Wang, Jie Peng, Tiancheng Huang, Sijia Guo, Huayang Wang, Zichen Zhu, Senyu Han, Yilu Cao, Kai Yu, Lu Chen, Kai Yu, and Lu Chen. Externalizing research synthesis and validation ...

2026 arXiv
[186]

A-evolve-training: Au- tonomous post-training of a 30b model, 2026

Zhan Shi, Bing He, Yisi Sang, Hanqing Lu, and Benoit Dumoulin. A-evolve-training: Au- tonomous post-training of a 30b model, 2026. URL https://arxiv.org/abs/2606.20657

2026 arXiv
[187]

One reflection is not enough: Self-correcting autonomous research via multi-hypothesis failure attribution, 2026

J Y, Binfei Chu, Gao, Jinlu Zhang, Yiwei Ma, Yi Tan, Jiayi Ji, Xiaoshuai Sun, and Rongrong Ji. One reflection is not enough: Self-correcting autonomous research via multi-hypothesis failure attribution, 2026. URL https://arxiv.org/abs/2606.31478

2026 arXiv
[188]

Baker, Benjamin Burns, 41 Daniel Adu-Ampratwum, Xuhui Huang, Xia Ning, Song Gao, Yu Su, and Huan Sun

Ziru Chen, Shijie Chen, Yuting Ning, Qianheng Zhang, Boshi Wang, Botao Yu, Yifei Li, Zeyi Liao, Chen Wei, Zitong Lu, Vishal Dey, Mingyi Xue, Frazier N. Baker, Benjamin Burns, 41 Daniel Adu-Ampratwum, Xuhui Huang, Xia Ning, Song Gao, Yu Su, and Huan Sun. Sci- enceAgentBench: To...

2024 arXiv
[189]

How far are we from true auto-research?, 2026

Zhengxin Zhang, Ning Wang, Sainyam Galhotra, and Claire Cardie. How far are we from true auto-research?, 2026. URL https://arxiv.org/abs/2605.19156

2026 arXiv
[190]

MLReplicate: Benchmarking autonomous research systems for machine learning reproducibil- ity, 2026

Sasi Kiran Gaddipati, Diyana Muhammed, Farhana Keya, Gollam Rabby, and Sören Auer. MLReplicate: Benchmarking autonomous research systems for machine learning reproducibil- ity, 2026. URL https://arxiv.org/abs/2605.16616

2026 arXiv
[191]

The calibration turn in AI-assisted research: A conceptual and methodological framework for evidence-licensed claims, 2026

Hongmin Li. The calibration turn in AI-assisted research: A conceptual and methodological framework for evidence-licensed claims, 2026. URL https://arxiv.org/abs/2606.31273

2026 arXiv
[192]

Grounded autonomous research: a fault-tolerant LLM pipeline from corpus to manuscript in frontier computational physics, 2026

H Huang. Grounded autonomous research: a fault-tolerant LLM pipeline from corpus to manuscript in frontier computational physics, 2026. URL https://arxiv.org/abs/2607.02329

2026 arXiv
[193]

Position: Correct answer, wrong mechanism – when AI scientists defend general claims their own data contradicts, 2026

Steven Young Eulig. Position: Correct answer, wrong mechanism – when AI scientists defend general claims their own data contradicts, 2026. URL https://arxiv.org/abs/2606.23175

2026 arXiv
[194]

Dead science walking: Publication bias and the AI scientist pipeline, 2026

Kargi Chauhan. Dead science walking: Publication bias and the AI scientist pipeline, 2026. URL https://arxiv.org/abs/2606.04220

2026 arXiv
[195]

(human) attention is (still) all you need: Human oversight makes AI-assisted social science reliable, 2026

Chen Zhu, Xiaolu Wang, and Weilong Zhang. (human) attention is (still) all you need: Human oversight makes AI-assisted social science reliable, 2026. URL https://arxiv.org/ab s/2606.12848

2026 arXiv
[196]

Boundless socratic learning with language games, 2024

Tom Schaul. Boundless socratic learning with language games, 2024. URL https://arxiv.or g/abs/2411.16905

2024 arXiv
[197]

Why self-rewarding works: Theoretical guarantees for iterative alignment of language models

Shi Fu, Yingjie Wang, Shengchao Hu, Peng Wang, and Dacheng Tao. Why self-rewarding works: Theoretical guarantees for iterative alignment of language models. Open MIND, 2026. doi: 10.48550/arxiv.2601.22513. URL https://arxiv.org/abs/2601.22513

2026 doi
[198]

The computational boundary of inference: Capability internalization, train- ing, and the turing jump, 2026

Chien-Ping Lu. The computational boundary of inference: Capability internalization, train- ing, and the turing jump, 2026. URL https://arxiv.org/abs/2605.27381

2026 arXiv
[199]

A mathematical framework for AI singularity: Conditions, bounds, and control of recursive improvement, 2025

Akbar Anbar Jafari, Cagrı Ozcınar, and Gholamreza Anbarjafari. A mathematical framework for AI singularity: Conditions, bounds, and control of recursive improvement, 2025. URL https://arxiv.org/abs/2511.10668

2025
[200]

Parker Whitfill and C.H. Wu. Will compute bottlenecks prevent an intelligence explosion?,
[201]

URL https://arxiv.org/abs/2507.23181

arXiv
[202]

Interestingness as an inductive heuristic for future compression progress, 2026

Vincent Herrmann and Jürgen Schmidhuber. Interestingness as an inductive heuristic for future compression progress, 2026. URL https://arxiv.org/abs/2605.14831

2026 arXiv
[203]

The unfireable safety kernel: Execution-time AI alignment for AI agents and other escapable AI systems, 2026

Seth Dobrin and Łukasz Chmiel. The unfireable safety kernel: Execution-time AI alignment for AI agents and other escapable AI systems, 2026. URL https://arxiv.org/abs/2606.26057

2026 arXiv
[204]

Inference scaling reshapes AI governance, 2025

Toby Ord. Inference scaling reshapes AI governance, 2025. URL https://arxiv.org/abs/2503 .05705

2025
[205]

Measuring biological capabil- ities and risks of AI agents, 2026

Patricia Paskov, Jeffrey Lee, Kyle Brady, and Alyssa Worland. Measuring biological capabil- ities and risks of AI agents, 2026. URL https://arxiv.org/abs/2606.19899. 42

2026 arXiv

Pith tools

Reviewed July 9, 2026 · model on record in the stance chip above.

[1] [27]

Adaptive self-improvement LLM agentic system for ML library development, 2025

Genghan Zhang, Weixin Liang, Olivia Hsu, and Kunle Olukotun. Adaptive self-improvement LLM agentic system for ML library development, 2025. URL https://arxiv.org/abs/2502.0 2534. 30

work page 2025

[2] [140]

Efficient Process Reward Modeling via Contrastive Mutual Information

Nakyung Lee, Sangwoo Hong, and Jungwoo Lee. Eﬀicient process reward modeling via contrastive mutual information, 2026. URL https://arxiv.org/abs/2604.10660

work page Pith review arXiv 2026

[3] [92]

Retrospective progress-aware self-refinement for LLM agent training, 2026

Xinbei Ma, Congmin Zheng, Jiyang Qiu, Jiale Hong, Yao Yao, Xiangmou Qu, Jiaxin Yin, Xingyu Lou, Jun Wang, Weiwen Liu, Weinan Zhang, Zhuosheng Zhang, and Hongyun Zhao. Retrospective progress-aware self-refinement for LLM agent training, 2026. URL https://ar xiv.org/abs/2606.14302

work page arXiv 2026

[4] [145]

Ali Heydari, Simon A

Weizhi Zhang, Zechen Li, Hamid Palangi, Ben Graef, A. Ali Heydari, Simon A. Lee, Sabera Rahman, Ray Luo, Zeinab Esmaeilpour, Erik Schenck, Chloe Zhang, Yamin Li, Menglian Zhou, Philip S. Yu, Daniel McDuff, Lindsey Sunden, Mark Malhotra, Shwetak Patel, and Ahmed A. Metwally. RubricsTree: Scalable and evolving open-ended evaluation of personal health agents...

work page 2026

[5] [175]

Effective Harness Engineering for Algorithm Discovery with Coding Agents

Yoichi Ishibashi, Taro Yano, and Masafumi Oyamada. Effective harness engineering for algorithm discovery with coding agents, 2026. URL https://arxiv.org/abs/2605.15221

work page Pith review arXiv 2026

[6] [5]

Recursive self-improvement

Anthropic. Recursive self-improvement. Anthropic Institute blog post, 2026. URL https: //www.anthropic.com/institute/recursive-self-improvement. Published May 2026

work page 2026

[7] [1]

Speculations concerning the first ultraintelligent machine

Irving John Good. Speculations concerning the first ultraintelligent machine. Advances in Computers, 6:31–88, 1966. doi: 10.1016/S0065-2458(08)60418-0

work page doi:10.1016/s0065-2458(08)60418-0 1966

[8] [2]

Goedel Machines: Self-Referential Universal Problem Solvers Making Provably Optimal Self-Improvements

Jürgen Schmidhuber. Gödel machines: Self-referential universal problem solvers making provably optimal self-improvements. arXiv preprint cs/0309048 , 2003. URL https://arxiv.or g/abs/cs/0309048

work page Pith review arXiv 2003