pith. sign in

arxiv: 2606.10934 · v1 · pith:J7YWELICnew · submitted 2026-06-09 · 💻 cs.AI

WorldKernel: A World Model is the Coupling Kernel of Admissible Possible Worlds

Fabio Rovai This is my paper

Pith reviewed 2026-06-27 12:59 UTC · model grok-4.3

classification 💻 cs.AI
keywords world modelscounterfactual reasoningstructural causal modelspositive semidefinite kernelscausal inferenceunidentified quantitiescoupling kernels
0
0 comments X

The pith

Prediction cannot represent uncertainty over counterfactual couplings between admissible worlds.

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

A strong predictor and Bayesian baseline both recover identified quantities from observational and interventional data across hundreds of structural causal models. On unidentified counterfactual couplings the predictor collapses to a point estimate, invalid in 28 percent of cases, while the truth remains an interval that more data never narrows. The gap arises because ordinary prediction recovers only the diagonal of the structure that determines counterfactuals. The paper models a world as one positive semidefinite coupling kernel whose off-diagonal entries encode the admissible cross-world couplings that every counterfactual query must read.

Core claim

A world model is the coupling kernel of admissible possible worlds: a single positive semidefinite kernel K(T,T') over pairs of admissible worlds whose diagonal recovers the ordinary posterior while the off-diagonal supplies the cross-world coupling information absent from marginal prediction and required by every counterfactual.

What carries the argument

The positive semidefinite coupling kernel K(T,T') over admissible worlds, whose diagonal is the posterior and whose off-diagonal encodes the admissible cross-world couplings.

If this is right

  • Positive semidefiniteness bounds counterfactual couplings in polynomial time where the exact response-type program remains intractable.
  • Ontology axioms tighten the resulting bounds by up to a third even on couplings they do not directly constrain.
  • Targeted scars learned from encountered infeasibilities close the gap several times faster than untargeted constraints.
  • Full reconstruction of the kernel reduces to approximate counting of admissible worlds, tractable below the Sly-Sun threshold.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Standard predictors may systematically output invalid counterfactual values on a substantial fraction of causal models even when given unlimited data.
  • Enforcing the kernel during world-model training could prevent collapse on unidentified queries without requiring full enumeration.
  • Decision systems that query counterfactuals may need to maintain and query the off-diagonal couplings explicitly rather than relying on marginal posteriors.

Load-bearing premise

Positive semidefiniteness of the coupling kernel supplies partial-identifying information about counterfactual couplings that is absent from the marginal posteriors alone.

What would settle it

Find two admissible worlds that share identical marginal posteriors yet differ on a cross-world counterfactual query, then check whether the kernel bound computed from positive semidefiniteness is violated by the true coupling.

Figures

Figures reproduced from arXiv: 2606.10934 by Fabio Rovai.

Figure 1
Figure 1. Figure 1: Permissive-lava arena. The unshielded planner reaches the goal by crossing the forbidden [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The off-diagonal witness. Both models share rungs 1 and 2 (the same two worlds [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Mediation X → M → Y . Fixing everything an experiment measures leaves the natural direct effect unidentified over an interval that spans zero. The endpoint models (blue) share identical rungs 1 and 2; the predictor (red) commits to one coupling. system direct effect (PN) mediation (NDE) reports the class? predictor (LLM) err 0.23; infeasible 28% not run no (point, often infeasible) diagonal-only (indep. SC… view at source ↗
Figure 4
Figure 4. Figure 4: SCM battery. Left: distribution of the diagonal-only error on the natural direct effect over 300 models, against the full-kernel oracle (exact). Right: the full kernel covers the truth on every model; the diagonal-only baseline is sign-unstable; the predictor returns infeasible answers on more than a quarter of models. Algorithm 2 Counterfactual bound from the kernel’s positive semidefiniteness (Prop. 3) R… view at source ↗
Figure 5
Figure 5. Figure 5: The kernel’s PSD structure as partial-identifying information, on the counterfactual [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: M suffices for second-order queries but not third. Left: on P(Y0=Y1=Y2=1) the exact response-type interval (green) is strictly inside the M-only bound (grey); the slack (red) is the untracked third moment E[Y0Y1Y2]. Right: the slack is a closed-form law, max(0, min(minij sij , mini di) − Pˆ 0) with Pˆ 0 = 1 − P i di + P ij sij , every random instance on the hinge to machine precision (positive on 24%); any… view at source ↗
Figure 7
Figure 7. Figure 7: Ontology-structured counterfactual bounds. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Online scarring. Left: blood-localized scars (the constraints the worst-case bound bleeds on) close the counterfactual identifiability gap up to 4× faster (early) than the same number of random constraints; a no-memory control stays at 0%. Right: Theorem 5 in numbers: the residual Fiedler value λ2 is exponentially small until targeted scarring removes the bottleneck phase, then jumps to Ω(1) (a threshold, … view at source ↗
Figure 9
Figure 9. Figure 9: Sly–Sun calipers. Left: the two coexisting phases are the caliper jaws, each a within-phase local computation (no global count). Right: the counterfactual is bracketed between the two phase-expectations EA[f], EB[f]; the true value (for the unknown phase weight α) lies inside, and the bracket width is that single unidentified scalar. 7 Intelligence as closure-preserving counterfactual competence Definition… view at source ↗
Figure 10
Figure 10. Figure 10: Where the calipers stop, measured. Left: few-phase regime, two modes near 1 that two probes resolve cleanly (phase split recovered at correlation 1.0). Right: glass, K near-degenerate modes crowd the top (gap ≈ 10−16); two probes capture two of K and underestimate the spread (2.0 vs true 7.0 at K = 8), and recovering all K needs a block of size K ∼ e Ω(n) . This is the Sly–Sun wall as a spectral phenomeno… view at source ↗
Figure 11
Figure 11. Figure 11: The counting transition, implemented. Left: the cavity order parameter (d − 1)η crosses 1 between degree 5 and 6, and belief-propagation error in counting independent sets explodes from 2% to 84% at the crossing. Right: the Fiedler value of the independent-set reconfiguration graph declines as the bottleneck tightens (the slow-mixing witness; the exp(−Ω(n)) collapse is asymptotic, so finite n shows a tren… view at source ↗
read the original abstract

A common assumption holds that enough observational and interventional data, given to a strong enough predictor, suffices. We report a failure mode that contradicts it. Across hundreds of structural causal models, on identified quantities a strong predictor and a Bayesian baseline both succeed, but on unidentified quantities (the couplings between counterfactual worlds) the predictor collapses to a point, on 28% of models to one no valid model can produce, while the truth is an admissible interval more data never narrows. The gap is structural: prediction cannot represent uncertainty over counterfactual couplings. We cast a world model as a single positive semidefinite coupling kernel K(T,T') over admissible worlds, whose diagonal is the ordinary posterior (what a predictor recovers) and whose off-diagonal is the cross-world coupling it cannot, which every counterfactual reads. The paper is the theory of that off-diagonal. It is real: two states with identical posteriors differ on a cross-world query, and the off-diagonal is the coupling that fixes counterfactuals. It can be bounded: positive semidefiniteness is partial-identifying information the marginals lack, and enforcing it bounds counterfactuals in polynomial time where the exact response-type program is intractable. Logical structure sharpens it: ontology axioms tighten the bound by up to a third, propagating to couplings they never touch. It can be acquired: targeted scars, constraints learned from encountered infeasibilities, close the gap several times faster than untargeted ones. Its full reconstruction is approximate counting of the admissible worlds, tractable below the Sly-Sun threshold and inapproximable above; we do not claim to beat the worst case.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper claims that standard predictors and Bayesian baselines succeed on identified quantities in structural causal models but fail on unidentified counterfactual couplings, collapsing to invalid point estimates on 28% of tested models while the truth is an admissible interval; it attributes this to a structural inability to represent uncertainty over cross-world couplings and proposes representing a world model as a positive semidefinite coupling kernel K(T,T') whose diagonal recovers the ordinary posterior and whose off-diagonals encode the missing couplings, with PSD supplying partial-identifying bounds enforceable in polynomial time (unlike exact response-type enumeration) and further tightened by ontology axioms.

Significance. If the central claims hold, the work would identify a previously unformalized representational gap in predictive models for counterfactual reasoning and supply a kernel-based mechanism for partial identification via PSD constraints, supported by the empirical observation of predictor failure across hundreds of SCMs. The computational claim of polynomial-time bounding would be a notable practical advantage if a compact formulation is provided.

major comments (3)
  1. [Abstract] Abstract: the claim that 'positive semidefiniteness is partial-identifying information the marginals lack, and enforcing it bounds counterfactuals in polynomial time where the exact response-type program is intractable' is load-bearing for the computational contribution, yet the abstract supplies neither an explicit SDP formulation, dual variables, nor any compact representation that would allow enforcement without constructing or optimizing over an explicitly exponential-sized kernel matrix indexed by admissible worlds.
  2. [Abstract] Abstract (and kernel construction): the assertion that the off-diagonal of K supplies information 'absent from the marginal posteriors alone' requires an explicit derivation or low-dimensional example (e.g., a 2x2 kernel) showing that the PSD constraint introduces independent bounds rather than being tautological with the definition of the admissible set; without this the partial-identification claim risks circularity.
  3. [Abstract] Abstract: the statement that 'full reconstruction is approximate counting of the admissible worlds, tractable below the Sly-Sun threshold' is used to contextualize the poly-time claim, but no argument is given for why the PSD enforcement procedure itself remains polynomial when the underlying counting problem is only approximable in restricted regimes.
minor comments (2)
  1. [Abstract] The term 'targeted scars' is introduced without a prior definition or reference to its formalization in the manuscript.
  2. [Abstract] The abstract refers to 'ontology axioms' tightening bounds by up to a third but does not indicate where in the manuscript these axioms are stated or how the propagation to untouched couplings is proved.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on the abstract. We believe the points raised can be addressed by revisions that improve clarity without altering the core contributions. We respond to each major comment in turn.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that 'positive semidefiniteness is partial-identifying information the marginals lack, and enforcing it bounds counterfactuals in polynomial time where the exact response-type program is intractable' is load-bearing for the computational contribution, yet the abstract supplies neither an explicit SDP formulation, dual variables, nor any compact representation that would allow enforcement without constructing or optimizing over an explicitly exponential-sized kernel matrix indexed by admissible worlds.

    Authors: We agree that the abstract would be improved by including more detail on the computational mechanism. The manuscript develops the bounding procedure via positive semidefiniteness constraints on the kernel; we will revise the abstract to reference the SDP formulation and dual variables for the marginal constraints as presented in the main text. revision: yes

  2. Referee: [Abstract] Abstract (and kernel construction): the assertion that the off-diagonal of K supplies information 'absent from the marginal posteriors alone' requires an explicit derivation or low-dimensional example (e.g., a 2x2 kernel) showing that the PSD constraint introduces independent bounds rather than being tautological with the definition of the admissible set; without this the partial-identification claim risks circularity.

    Authors: The main text contains a low-dimensional example with a 2x2 kernel over two admissible worlds that share the same marginal posterior but differ in their cross-world coupling, where the PSD constraint yields strictly tighter bounds. We will incorporate a brief version of this example into the revised abstract to demonstrate the independent information supplied by the off-diagonal. revision: yes

  3. Referee: [Abstract] Abstract: the statement that 'full reconstruction is approximate counting of the admissible worlds, tractable below the Sly-Sun threshold' is used to contextualize the poly-time claim, but no argument is given for why the PSD enforcement procedure itself remains polynomial when the underlying counting problem is only approximable in restricted regimes.

    Authors: The polynomial-time bounding applies to the SDP over the kernel matrix dimension, while approximate counting pertains only to full kernel reconstruction. We will revise the abstract to clarify this distinction and note that the bounding procedure operates directly on the PSD and marginal constraints. revision: yes

Circularity Check

1 steps flagged

Kernel definition includes off-diagonal coupling by construction

specific steps
  1. self definitional [abstract]
    "We cast a world model as a single positive semidefinite coupling kernel K(T,T') over admissible worlds, whose diagonal is the ordinary posterior (what a predictor recovers) and whose off-diagonal is the cross-world coupling it cannot, which every counterfactual reads. ... positive semidefiniteness is partial-identifying information the marginals lack"

    The construction defines the world model to be exactly the object that possesses the off-diagonal couplings standard prediction lacks; the claim that PSD supplies identifying information absent from marginal posteriors is therefore true by the definition of K rather than shown to follow from it.

full rationale

The paper's central move defines a world model directly as a PSD kernel whose off-diagonal entries are the cross-world couplings absent from ordinary posteriors. This matches the self-definitional pattern: the claimed structural gap and its partial identification via PSD are introduced as part of the object's definition rather than derived from independent premises or data. No other load-bearing steps (self-citations, fitted predictions, or ansatzes) are quotable from the supplied text as reducing to inputs. The poly-time enforcement claim is noted but lacks an explicit equation or reduction showing it collapses to the definition.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

Review performed on abstract only; full text unavailable so ledger entries are limited to concepts explicitly named in the abstract.

axioms (2)
  • domain assumption Positive semidefiniteness supplies partial-identifying information absent from marginal posteriors
    Abstract states this is the source of bounds on counterfactuals
  • domain assumption Ontology axioms tighten bounds on couplings they never touch
    Abstract claims propagation occurs without direct contact
invented entities (1)
  • Coupling kernel K(T,T') no independent evidence
    purpose: Encodes both posterior on diagonal and cross-world couplings on off-diagonal for admissible worlds
    New construct introduced to represent the information standard prediction cannot recover

pith-pipeline@v0.9.1-grok · 5822 in / 1442 out tokens · 23385 ms · 2026-06-27T12:59:49.509795+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

59 extracted references · 7 linked inside Pith

  1. [1]

    Duarte, N

    G. Duarte, N. Finkelstein, D. Knox, J. Mummolo, I. Shpitser. An Automated Approach to Causal Inference in Discrete Settings. Journal of the American Statistical Association, 2024

    2024

  2. [2]

    Assran et al

    M. Assran et al. V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning. arXiv:2506.09985, 2025

    Pith/arXiv arXiv 2025

  3. [3]

    Cosmos 3: The Open Physical AI Foundation Model

    NVIDIA. Cosmos 3: The Open Physical AI Foundation Model. Technical report, 2026

    2026

  4. [4]

    Marble: A Multimodal World Model

    World Labs. Marble: A Multimodal World Model. 2025

    2025

  5. [5]

    J. Xu, Z. Zhang, T. Friedman, Y. Liang, G. Van den Broeck. A Semantic Loss Function for Deep Learning with Symbolic Knowledge. ICML, 2018

    2018

  6. [6]

    Alshiekh, R

    M. Alshiekh, R. Bloem, R. Ehlers, B. K¨ onighofer, S. Niekum, U. Topcu. Safe Reinforcement Learning via Shielding. AAAI, 2018

    2018

  7. [7]

    A. D. Ames et al. Control Barrier Functions: Theory and Applications. ECC, 2019

    2019

  8. [8]

    L. Chlon. NTK-Mirror: LoRA-free Forward-Pass Fine-Tuning via Signed Log-Mask Controllers. Software repository, 2026.https://github.com/leochlon/ntkmirror

    2026

  9. [9]

    P. C. G. da Costa, K. B. Laskey, K. J. Laskey. PR-OWL: A Bayesian Ontology Language for the Semantic Web. URSW, 2005

    2005

  10. [10]

    De Raedt, A

    L. De Raedt, A. Kimmig, H. Toivonen. ProbLog: A Probabilistic Prolog. IJCAI, 2007

    2007

  11. [11]

    Jerrum, L

    M. Jerrum, L. Valiant, V. Vazirani. Random Generation of Combinatorial Structures from a Uniform Distribution. TCS, 1986

    1986

  12. [12]

    Aharonov, A

    D. Aharonov, A. Ta-Shma. Adiabatic Quantum State Generation and Statistical Zero Knowledge. STOC, 2003

    2003

  13. [13]

    Grover, T

    L. Grover, T. Rudolph. Creating Superpositions that Correspond to Efficiently Integrable Probability Distributions. arXiv:quant-ph/0208112, 2002

    Pith/arXiv arXiv 2002

  14. [14]

    D. Weitz. Counting Independent Sets up to the Tree Threshold. STOC, 2006

    2006

  15. [15]

    A. Sly. Computational Transition at the Uniqueness Threshold. FOCS, 2010

    2010

  16. [16]

    A. Sly, N. Sun. Counting in Two-Spin Models on d-Regular Graphs. Annals of Probability, 2014

    2014

  17. [17]

    J. Pearl. Causality: Models, Reasoning, and Inference. Cambridge University Press, 2009

    2009

  18. [18]

    J. Tian, J. Pearl. Probabilities of Causation: Bounds and Identification. Annals of Mathematics and AI, 2000

    2000

  19. [19]

    F. Rovai. Open Ontologies: Tool-Augmented Ontology Engineering with Stable Matching Alignment. arXiv:2605.09184, 2026

    Pith/arXiv arXiv 2026

  20. [20]

    F. Rovai. CIVeX: Causal Intervention Verification for Language Agents. arXiv:2605.09168, 2026. 24

    Pith/arXiv arXiv 2026

  21. [21]

    F. Rovai. Deterministic Event-Graph Substrates as World Models for Counterfactual Reasoning. arXiv:2605.15967, 2026

    Pith/arXiv arXiv 2026

  22. [22]

    F. Rovai. Saturating Scaling Laws for Equational Discovery. arXiv:2605.23983, 2026

    Pith/arXiv arXiv 2026

  23. [23]

    D. Ha, J. Schmidhuber. Recurrent World Models Facilitate Policy Evolution. NeurIPS, 2018

    2018

  24. [24]

    Hafner, J

    D. Hafner, J. Pasukonis, J. Ba, T. Lillicrap. Mastering Diverse Control Tasks through World Models (DreamerV3). Nature, 2025

    2025

  25. [25]

    Y. LeCun. A Path Towards Autonomous Machine Intelligence. OpenReview, 2022

    2022

  26. [26]

    Bruce et al

    J. Bruce et al. Genie: Generative Interactive Environments. ICML, 2024

    2024

  27. [27]

    Kıcıman, R

    E. Kıcıman, R. Ness, A. Sharma, C. Tan. Causal Reasoning and Large Language Models: Opening a New Frontier for Causality. arXiv:2305.00050, 2023

    arXiv 2023

  28. [28]

    Zeˇ cevi´ c, M

    M. Zeˇ cevi´ c, M. Willig, D. S. Dhami, K. Kersting. Causal Parrots: Large Language Models May Talk Causality But Are Not Causal. TMLR, 2023

    2023

  29. [29]

    Jin et al

    Z. Jin et al. CLadder: Assessing Causal Reasoning in Language Models. NeurIPS, 2023

    2023

  30. [30]

    Jin et al

    Z. Jin et al. Can Large Language Models Infer Causation from Correlation? ICLR, 2024

    2024

  31. [31]

    Y. Chen, V. K. Singh, J. Ma, R. Tang. CounterBench: Evaluating and Improving Counterfactual Reasoning in LLMs. arXiv:2502.11008, 2025

    Pith/arXiv arXiv 2025

  32. [32]

    K. Vafa, J. Y. Chen, A. Rambachan, J. Kleinberg, S. Mullainathan. Evaluating the World Model Implicit in a Generative Model. NeurIPS, 2024

    2024

  33. [33]

    K. Vafa, P. G. Chang, A. Rambachan, S. Mullainathan. What Has a Foundation Model Found? Using Inductive Bias to Probe for World Models. ICML, 2025

    2025

  34. [34]

    E. M. Bender, T. Gebru, A. McMillan-Major, S. Shmitchell. On the Dangers of Stochastic Parrots. FAccT, 2021

    2021

  35. [35]

    Bareinboim, J

    E. Bareinboim, J. D. Correa, D. Ibeling, T. Icard. On Pearl’s Hierarchy and the Foundations of Causal Inference. In Probabilistic and Causal Inference, ACM Books, 2022

    2022

  36. [36]

    Ibeling, T

    D. Ibeling, T. Icard. Probabilistic Reasoning across the Causal Hierarchy. AAAI, 2020

    2020

  37. [37]

    Xia, K.-Z

    K. Xia, K.-Z. Lee, Y. Bengio, E. Bareinboim. The Causal-Neural Connection: Expressiveness, Learnability, and Inference. NeurIPS, 2021

    2021

  38. [38]

    J. D. Correa, S. Lee, E. Bareinboim. Nested Counterfactual Identification from Arbitrary Surrogate Experiments. NeurIPS, 2021

    2021

  39. [39]

    Zhang, J

    J. Zhang, J. Tian, E. Bareinboim. Partial Counterfactual Identification from Observational and Experimental Data. ICML, 2022

    2022

  40. [40]

    A. Li, J. Pearl. Probabilities of Causation: Role of Observational Data. AISTATS, 2023

    2023

  41. [41]

    C. F. Manski. Nonparametric Bounds on Treatment Effects. American Economic Review, 80(2):319–323, 1990. 25

    1990

  42. [42]

    Balke, J

    A. Balke, J. Pearl. Counterfactual Probabilities: Computational Methods, Bounds and Appli- cations. UAI, 1994

    1994

  43. [43]

    J. M. Robins, S. Greenland. Identifiability and Exchangeability for Direct and Indirect Effects. Epidemiology, 3(2):143–155, 1992

    1992

  44. [44]

    J. Pearl. Direct and Indirect Effects. UAI, 2001

    2001

  45. [45]

    C. Avin, I. Shpitser, J. Pearl. Identifiability of Path-Specific Effects. IJCAI, 2005

    2005

  46. [46]

    K. Imai, L. Keele, T. Yamamoto. Identification, Inference and Sensitivity Analysis for Causal Mediation Effects. Statistical Science, 25(1):51–71, 2010

    2010

  47. [47]

    T. J. VanderWeele. Explanation in Causal Inference: Methods for Mediation and Interaction. Oxford University Press, 2015

    2015

  48. [48]

    J. M. Robins, T. S. Richardson. Alternative Graphical Causal Models and the Identification of Direct Effects. In Causality and Psychopathology, Oxford University Press, 2010

    2010

  49. [49]

    R. M. Andrews, V. Didelez. Insights into the Cross-World Independence Assumption of Causal Mediation Analysis. Epidemiology, 32(2):209–219, 2021

    2021

  50. [50]

    A. P. Dawid. Causal Inference without Counterfactuals. JASA, 95(450):407–424, 2000

    2000

  51. [51]

    Heckerman, R

    D. Heckerman, R. Shachter. Decision-Theoretic Foundations for Causal Reasoning. JAIR, 3:405–430, 1995

    1995

  52. [52]

    Liang et al

    Y. Liang et al. VisualPredicator: Learning Abstract World Models with Neuro-Symbolic Predicates for Robot Planning. ICLR, 2025

    2025

  53. [53]

    Manhaeve, S

    R. Manhaeve, S. Dumanˇ ci´ c, A. Kimmig, T. Demeester, L. De Raedt. DeepProbLog: Neural Probabilistic Logic Programming. NeurIPS, 2018

    2018

  54. [54]

    Richardson, P

    M. Richardson, P. Domingos. Markov Logic Networks. Machine Learning, 62:107–136, 2006

    2006

  55. [55]

    Riguzzi, E

    F. Riguzzi, E. Bellodi, E. Lamma, R. Zese. Probabilistic Description Logics under the Distribu- tion Semantics. Semantic Web, 6(5):477–501, 2015

    2015

  56. [56]

    Chavira, A

    M. Chavira, A. Darwiche. On Probabilistic Inference by Weighted Model Counting. Artificial Intelligence, 172(6–7):772–799, 2008

    2008

  57. [57]

    Darwiche, P

    A. Darwiche, P. Marquis. A Knowledge Compilation Map. JAIR, 17:229–264, 2002

    2002

  58. [58]

    L. G. Valiant. The Complexity of Computing the Permanent. Theoretical Computer Science, 8(2):189–201, 1979

    1979

  59. [59]

    Galanis, D

    A. Galanis, D. ˇStefankoviˇ c, E. Vigoda. Inapproximability of the Partition Function for the Antiferromagnetic Ising and Hard-Core Models. Combinatorics, Probability and Computing, 25(4):500–559, 2016. 26

    2016