# Stefan Edelkamp: Inner-Learning in Trick-Taking Games [Canonical reading](https://escapemeridian.com/readings/stefan-edelkamp-inner-learning-in-trick-taking-games-290f9429) Stefan Edelkamp presents a declarative approach to belief-state modeling in trick-taking card games, starting with a state-of-the-art Skat player. The work translates the player’s hand-coded inference rules into a domain-specific language, then adds rules derived by symmetry and annotated from an expert strategy document. It also represents cardinality constraints—bounds on how many cards a player holds from a named group—and enforces them when generating possible game worlds. Confidence parameters connect the symbolic rules to probabilistic belief updates. The abstract says round-trip code generation checks whether the extracted rules reproduce the original engine’s intended behavior; the declarative engine also corrects up to three audited bugs in that engine. In self-play evaluation, Edelkamp reports improvements in winning percentage and score over the original hand-coded engine, with extracted rules, human-annotated rules and cardinality constraints each contributing. Escape Meridian’s reading: this is a concrete demonstration in trick-taking games of combining existing code, structured rules and expert knowledge in one representation. The abstract does not give the size of the measured gains or enough evaluation detail to assess how broadly they apply. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source arXiv cs.AI · Article Original source date: Oct 28, 2026 (2026-10-28). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-11T13:22:43.071Z. First observed: 2026-10-30T06:42:52.210Z. Retrieved: 2026-10-31T05:16:51.396Z. - [Inner-Learning: From Hand-Coded Beliefs to Declarative Knowledge in Trick-Taking Games](https://arxiv.org/abs/2610.38304) ## Limits The supplied evidence is the arXiv abstract; it does not give numerical gain sizes or detailed self-play evaluation conditions. The reported evaluation concerns trick-taking card games; the abstract does not establish whether the approach transfers to other settings. ## Open question How large were the reported self-play improvements, and under what evaluation conditions were they measured? ## Source-stated Edelkamp presents a declarative approach to belief-state modeling for trick-taking card games, beginning with a state-of-the-art Skat player. > We present a declarative approach to belief-state modeling in trick-taking card games. Starting from a state-of-the-art Skat player - [Supporting source](https://arxiv.org/abs/2610.38304) ## Source-stated The system extracts hand-coded inference rules into a domain-specific language and checks that generated code reproduces the original engine’s intended behavior, while correcting up to three audited bugs. > Round-trip code generation verifies that the extracted rules reproduce the original engine's intended behavior, up to three audited bugs in the original that the declarative engine corrects. - [Supporting source](https://arxiv.org/abs/2610.38304) ## Source-stated The authors extend the rule base using symmetric derivation and annotations from an expert strategy document, and add cardinality constraints enforced during world generation. > We then grow the rule base through symmetric derivation and annotation of an expert strategy document---unifying code, symmetry, and expertise in one representation. We extend the knowledge representation with \\emph{cardinality constraints} - [Supporting source](https://arxiv.org/abs/2610.38304) ## Source-stated The abstract reports that self-play evaluation improved winning percentage and score over the original hand-coded engine, with each of three rule sources contributing. > Self-play evaluation shows the declarative engine measurably improving winning percentage and score over the original hand-coded engine, with the extracted rules, the human-annotated rules, and the cardinality constraints each contributing to the effect. - [Supporting source](https://arxiv.org/abs/2610.38304) ## Escape Meridian interpretation The abstract does not report the magnitude of the gains or provide the evaluation details needed to judge their scope. > Self-play evaluation shows the declarative engine measurably improving winning percentage and score over the original hand-coded engine - [Supporting source](https://arxiv.org/abs/2610.38304) ## Source and contributor context Edelkamp, Stefan · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 5c72e09c-541e-4e2b-8d73-f7a9e1f0b8c9. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-11T13:22:43.071Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 6f16232e488fdec519d05949a6f70fd001117eef6900494082a77d11e8108485. Source text hash: f4592914e91de18049c87b8237f77ebca3d905ac37bfa53938c8f34d7bf91c92. ## Revision and source-check history ### Version 1 — 2026-11-11T13:22:43.071Z Stefan Edelkamp presents a declarative approach to belief-state modeling in trick-taking card games, starting with a state-of-the-art Skat player. The work translates the player’s hand-coded inference rules into a domain-specific language, then adds rules derived by symmetry and annotated from an expert strategy document. It also represents cardinality constraints—bounds on how many cards a player holds from a named group—and enforces them when generating possible game worlds. Confidence parameters connect the symbolic rules to probabilistic belief updates. The abstract says round-trip code generation checks whether the extracted rules reproduce the original engine’s intended behavior; the declarative engine also corrects up to three audited bugs in that engine. In self-play evaluation, Edelkamp reports improvements in winning percentage and score over the original hand-coded engine, with extracted rules, human-annotated rules and cardinality constraints each contributing. Escape Meridian’s reading: this is a concrete demonstration in trick-taking games of combining existing code, structured rules and expert knowledge in one representation. The abstract does not give the size of the measured gains or enough evaluation detail to assess how broadly they apply. ## Original source arXiv cs.AI · Article Original source date: Oct 28, 2026 (2026-10-28). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-11T13:22:43.071Z. First observed: 2026-10-30T06:42:52.210Z. Retrieved: 2026-10-31T05:16:51.396Z. - [Inner-Learning: From Hand-Coded Beliefs to Declarative Knowledge in Trick-Taking Games](https://arxiv.org/abs/2610.38304) ## Limits The supplied evidence is the arXiv abstract; it does not give numerical gain sizes or detailed self-play evaluation conditions. The reported evaluation concerns trick-taking card games; the abstract does not establish whether the approach transfers to other settings. ## Open question How large were the reported self-play improvements, and under what evaluation conditions were they measured? ## Source-stated Edelkamp presents a declarative approach to belief-state modeling for trick-taking card games, beginning with a state-of-the-art Skat player. > We present a declarative approach to belief-state modeling in trick-taking card games. Starting from a state-of-the-art Skat player - [Supporting source](https://arxiv.org/abs/2610.38304) ## Source-stated The system extracts hand-coded inference rules into a domain-specific language and checks that generated code reproduces the original engine’s intended behavior, while correcting up to three audited bugs. > Round-trip code generation verifies that the extracted rules reproduce the original engine's intended behavior, up to three audited bugs in the original that the declarative engine corrects. - [Supporting source](https://arxiv.org/abs/2610.38304) ## Source-stated The authors extend the rule base using symmetric derivation and annotations from an expert strategy document, and add cardinality constraints enforced during world generation. > We then grow the rule base through symmetric derivation and annotation of an expert strategy document---unifying code, symmetry, and expertise in one representation. We extend the knowledge representation with \\emph{cardinality constraints} - [Supporting source](https://arxiv.org/abs/2610.38304) ## Source-stated The abstract reports that self-play evaluation improved winning percentage and score over the original hand-coded engine, with each of three rule sources contributing. > Self-play evaluation shows the declarative engine measurably improving winning percentage and score over the original hand-coded engine, with the extracted rules, the human-annotated rules, and the cardinality constraints each contributing to the effect. - [Supporting source](https://arxiv.org/abs/2610.38304) ## Escape Meridian interpretation The abstract does not report the magnitude of the gains or provide the evaluation details needed to judge their scope. > Self-play evaluation shows the declarative engine measurably improving winning percentage and score over the original hand-coded engine - [Supporting source](https://arxiv.org/abs/2610.38304) ## Source and contributor context Edelkamp, Stefan · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 5c72e09c-541e-4e2b-8d73-f7a9e1f0b8c9. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-11T13:22:43.071Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 6f16232e488fdec519d05949a6f70fd001117eef6900494082a77d11e8108485. Source text hash: f4592914e91de18049c87b8237f77ebca3d905ac37bfa53938c8f34d7bf91c92. ### uncertain — 2026-11-11T13:22:43.071Z — anthropic/claude-opus-5.5 Publication: 5c72e09c-541e-4e2b-8d73-f7a9e1f0b8c9 attribution — pass: The work is correctly attributed to sole author Stefan Edelkamp via arXiv, and the claims stay attributed to the abstract. One claim says 'The authors extend', while the source lists a single author. This is a minor wording slip, not misattribution; the smallest repair is 'Edelkamp extends'. > Authors:Stefan Edelkamp summary — uncertain: The reading is faithful in structure but hard for a non-specialist to follow. Central terms are left unexplained in plain language: belief-state modeling (tracking what an AI infers about opponents' hidden cards), declarative, domain-specific language, round-trip code generation, self-play and Skat. The title term Inner-Learning is also never explained. The source's own primitive, that a player cannot hold any card in a set, would make the idea concrete but is omitted. Smallest repair: one plain-language gloss of belief states with that primitive, and a note that Skat is a card game. Term-library tooltips should be added for the listed terms. > denoting that player $P$ cannot hold any card in set $C$ date — pass: The originalPublished date of 2026-10-28 matches the arXiv v1 submission date. It is not the later feed date of 2026-10-31. > \[Submitted on 28 Oct 2026\] support — uncertain: In the source, 'up to three audited bugs' means the extracted rules reproduce the original's intended behavior except for three audited bugs, which the declarative engine corrects. The summary instead says the engine 'corrects up to three audited bugs', which reads as an upper bound and could imply fewer than three were fixed. Smallest repair: 'apart from three audited bugs in the original, which the declarative engine corrects'. The other claims are source-stated. The Escape Meridian interpretation is labeled, and the stated gap about gain size is accurate. > reproduce the original engine's intended behavior, up to three audited bugs in the original that the declarative engine corrects media — pass: The selected media is the verified direct arXiv abstract page for this paper. It is the most direct access to the source. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/290f9429096dec4ef971f53e4654023cae73d56519e3266b9eec9c72e96aefff/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/stefan-edelkamp-inner-learning-in-trick-taking-games-290f9429.json)