Abstraction Agent

Boning Li; Longbo Huang

cs.MASep 3, 2026

Abstraction Agent

Authors: Boning Li, Longbo Huang

Organizations: Institute for Interdisciplinary Information Sciences, Tsinghua University

Abstract

Information abstraction, which groups strategically similar private states into a tractable number of buckets, is essential for scaling game-solving algorithms to large imperfect-information games. Constructing effective abstractions, however, has traditionally required domain-specific evaluators such as hand-strength calculators or equity estimators, which demand expert knowledge and engineering effort and are unavailable for most less-studied games. We propose the Abstraction Agent, a zero-shot pipeline that uses a large language model (LLM) to discover continuous strategic features from a natural-language game description, score private states on these features, and cluster them into abstraction buckets, without any game-specific evaluator, training data, or game-tree traversal during abstraction construction. The pipeline runs in four phases: feature discovery with calibration anchors, batched private-state scoring, correlation-based feature selection, and $k$ -means clustering. The resulting abstractions reduce lifted-strategy exploitability by up to 62% relative to an expected-hand-strength baseline on heads-up no-limit Texas hold'em (HUNL) turn endgames, and beat a scalar rank baseline at every granularity on ROVER Trials, an original game absent from any pretraining corpus. Beyond these quantitative benchmarks, the pipeline transfers with unchanged prompts to four-card Pot-Limit Omaha, HUNL preflop and flop, and Riichi Mahjong, where the discovered features track each game's recognized strategic concepts. This is structured knowledge elicitation: converting implicit strategic knowledge in LLM parameters into explicit numerical features for downstream algorithmic computation. The code is available at https://github.com/lbn187/AbstractionAgent.

Figures & tables

Figure 1: The Abstraction Agent pipeline. Given a textual GameConfig and a set of private holdings, an LLM discovers strategic features (Phase 1), scores each holding on them (Phase 2), redundant dimensions are filtered (Phase 3), and holdings are clustered into

K

buckets (Phase 4). The pipeline uses no hand evaluators, training data, or game-tree traversal, and the same code applies to poker, Omaha, and Mahjong unchanged.

Board	$K$	Agent	EHS	Ratio
7 $\spadesuit$ 9 $\heartsuit$ 9 $\clubsuit$ T $\clubsuit$	20	0.0437	0.0335	1.30 $\pm$ 0.20
	50	0.0142	0.0192	0.74 $\pm$ 0.17
	100	0.0036	0.0098	0.36 $\pm$ 0.04
T $\spadesuit$ 6 $\heartsuit$ A $\heartsuit$ 7 $\clubsuit$	20	0.1359	0.1088	1.25 $\pm$ 0.23
	50	0.0483	0.0592	0.82 $\pm$ 0.10
	100	0.0102	0.0256	0.40 $\pm$ 0.08

Table 1: HUNL turn endgame exploitability (fraction of pot; mean over 10 seeds). Ratio = mean Agent / EHS

\pm

95% Student-

t

CI half-width.

$K$	Backbone	SG1	SG2
50	GPT-5.5	0.74	0.82
	Llama-3.3-70B	1.62	1.86
	Qwen-2.5-32B	1.48	2.01
100	GPT-5.5	0.36	0.40
	Llama-3.3-70B	1.07	1.36
	Qwen-2.5-32B	1.96	3.70

Table 2: Backbone dependence: mean exploitability ratio (Agent / EHS) over 10 seeds.

Domain-specific methods (require hand evaluator):
Method	SG1	SG2
Random	0.1314	0.3670
LLM-Embed (zero-shot)	0.0318	0.1847
Abstraction Agent (ours)	0.0142	0.0483
EHS	0.0192	0.0592
PA (EHS + Variance)	0.0053	0.0254

Table 3: Comparison of abstraction methods at

K{=}50

(exploitability as fraction of pot; agent is the 10-seed mean).

Backbone	$K$	Agent/Rank	Agent/PA
GPT-5.5	5	0.920 $\pm$ 0.074	1.043 $\pm$ 0.106
	10	0.804 $\pm$ 0.038	0.940 $\pm$ 0.067
	20	0.667 $\pm$ 0.070	0.913 $\pm$ 0.106
	30	0.663 $\pm$ 0.065	0.916 $\pm$ 0.111
Qwen-2.5-32B	5	1.157 $\pm$ 0.085	1.319 $\pm$ 0.160
	10	0.969 $\pm$ 0.059	1.128 $\pm$ 0.054

Table 4: ROVER Trials: mean exploitability ratio (Agent / baseline) over 10 seeds

\pm

95% Student-

t

CI half-width.

Appendix figures & tables13 assets

Supplementary material from the paper’s appendix.

Appendix

Pattern	Count	Description
Double-suited (2+2)	3,081	Two suited pairs, two flush draws.
Single-suited (2+1+1)	7,098	One suited pair; one flush draw.
Triple-suited (3+1)	3,718	Three of one suit, one weaker flush draw.
Rainbow (1+1+1+1)	1,820	All four suits differ; no flush potential.
Monotone (4+0)	715	Same suit; one flush draw (must-use-two).

Appendix

Table 5: PLO4 suit pattern classification (canonical hand counts out of 16,432).

Type	Example	Strategic Properties
Premium pairs + suited	AA-KK ds	Top pair + flush draws; strongest
Top rundowns	AKQJ, KQJT	Max straight connectivity
Suited connectors	8h7h6s5s	Flush + straight; strong playability
High pairs + danglers	AA-xx rainbow	High pair, limited coordination
Mid rundowns	T987, 9876	Good connectivity, lower nut potential
Suited aces	As-xxx	Nut flush draw potential

Appendix

Table 6: PLO4 structural hand types with examples and strategic properties.

ID	Board	Street	Pot
1	$7\spadesuit\,9\heartsuit\,9\clubsuit\,T\clubsuit$	Turn	500
2	$T\spadesuit\,6\heartsuit\,A\heartsuit\,7\clubsuit$	Turn	4780

Appendix

Table 7: Libratus subgame specifications.

Game	$N$	$K$	Key Result
HUNL Turn	${\sim}$ 1K	20–100	up to 62% $\downarrow$ vs EHS
ROVER Trials	125	5–30	beats baseline
PLO4 Preflop	16,432	30	PLO features
HUNL Flop	1,176	20	Draw/made sep.
HUNL Preflop	169	10	Premium groups
Mahjong	200	10	Chance features

Appendix

Table 8: Cross-game summary of the six evaluation settings.

Cluster	Representative Hands	Characteristics
0 (8)	22, 33, 44, 55, 66, 77, 88, 99	Small/medium pairs (set-mining value)
1 (19)	A2s–A5s, 54s–76s	Suited aces + suited connectors
2 (19)	32o–T2o	Offsuit junk (weakest hands)
3 (12)	QTo, KTo, ATo, QJo, KJo, AJo	Offsuit broadways
4 (25)	K2s–K4s, Q2s–Q4s, T4s–J4s	Suited high + low kicker
5 (10)	ATs, QJs, KJs, AJs, KQs	Premium suited broadways

Appendix

Table 9: Complete HUNL preflop clusters (

K=10

, 169 canonical hands, sorted by ID).

Feature	SG1	SG2
Current_Made_Str.	0.87	0.93
Improvement_Pot.	0.27	$-$ 0.30
Vuln._To_Outdraws	0.68	0.31
Nut_Ceiling	0.68	0.14
River_Stability	$-$ 0.36	0.52
Blocker_Value	0.65	0.46

Appendix

Table 10: Pearson correlation between LLM features and EHS across turn subgames.

Cluster	Representative Hands	Characteristics
0 (590)	9 a 7 b 5 a 5 b (ds), 6 a 6 b 5 a 3 b (ds)	Low-rank suited coordination holdings
1 (505)	A a J b J c 7 b (ss), A a J b J c 6 b (ss)	High-pair heavy
2 (629)	K a T b 9 a 5 b (ds), A a T a 5 b 2 b (ds)	Low-rank suited coordination holdings
3 (324)	A a A b K c 5 c (ss), A a A b 9 c 7 c (ss)	Premium paired / AA-heavy holdings
4 (743)	Q a 9 a 6 b 3 b (ds), J a T b 7 a 3 a (ts)	Low-rank suited coordination holdings
5 (468)	Q a T b 9 c 6 a (ss), Q a J b 8 a 8 c (ss)	Mixed suited coordination holdings

Appendix

Table 11: Complete PLO4 preflop clusters (

K=30

, 16,432 canonical hands; abstract suit labels

a

–

d

Cluster	Representative Hands	Characteristics
16 (133)	J a 9 b 7 c 5 d (r), Q a T b 9 c 7 d (r)	Rainbow low-to-mid coordination holdings
17 (178)	K a K b T c 4 d (r), A a A b Q c 7 d (r)	Premium paired / AA-heavy holdings
18 (624)	Q a J a 8 b 2 a (ts), 5 a 5 b 2 a 2 b (ds)	Low-rank suited coordination holdings
19 (387)	7 a 5 b 2 c 2 d (r), A a 7 b 3 c 2 d (r)	Rainbow low-to-mid coordination holdings
20 (896)	9 a 7 a 2 a 2 b (ts), Q a 7 a 7 b 2 a (ts)	Low-pair heavy
21 (618)	A a K b 9 a 9 b (ds), K a Q b T b 2 a (ds)	Mixed suited coordination holdings

Appendix

Table 12: PLO4 preflop clusters (continued).

Cluster	Representative Hands	Characteristics
0 (80)	Ad 9s, Ad 9d, Ad 8c	Top pair A, strong kicker
1 (16)	Js Ts, Js Td, Jd Tc	Nut straight (Broadway)
2 (162)	9s 4d, 9s 3c, 9s 2h	Air (9x low kicker, no draw)
3 (138)	Ah 5h, Ah 4h, Ks 9s	Medium made + backdoor draws
4 (25)	9d 9c, 8s 8h, 6s 6d	Underpairs (below board)
5 (8)	Ad Ac, Ks Kd, Qs Qc	Sets (AA/KK/QQ)

Appendix

Table 13: Complete HUNL flop clusters (board A

\spadesuit

\heartsuit

\diamondsuit

;

K=20

, 1,176 hands, sorted by ID).

Variant	$K{=}20$		$K{=}50$
Variant	SG1	SG2	SG1	SG2
Rules-only	1.20	1.21	0.73	0.91
Range-aware	1.30	1.25	0.74	0.82
Feature-discovery	1.35	1.18	0.76	0.92

Appendix

Table 14: Prompt-variant robustness: mean Agent / EHS exploitability ratio over 10 seeds (GPT-5.5).

Feature Removed	Mean ARI
Vulnerability_To_Outdraws	0.7558
Blocker_Value	0.7607
Current_Made_Strength	0.7638
Nut_Ceiling	0.7716
Improvement_Potential	0.7866
River_Stability	0.7959

Appendix

Table 15: Feature importance via leave-one-out ablation (mean ARI; lower = larger impact).

Feature	Description
Current_Speed	Shanten + effective tiles
Acceptance_Breadth	Tile types improving hand
Hand_Value_Potential	Scoring upside (yaku, dora)
Wait_Quality	Quality of the winning wait
Defensive_Resilience	Avoid dealing into opponents
Next_Draw_Swing	How one draw changes value (not selected)

Appendix

Table 16: Features discovered for Riichi Mahjong mid-game (turn 8, East round).

Size	Representative Hands	Dominant Trait
27	11578m 156p 13347s GR	Wide acceptance, 3–4 shanten
20	112467899m 4p 16s W	High speed + value
22	11149m 5579p 668s WH	Slow, high defense
16	1129m 39p 4777s WGR	High value, slow
23	1136m 2578p 1455s S	Average hands
33	1119m 1p 157889s WR	Most defensive

Appendix

Table 17: Complete Riichi Mahjong clusters (

K=10

, 200 hands).

Imperfect-Information Games

Imperfect-information games study strategic decisions when players have private information or cannot fully observe the state or actions.

See all 90 papers

May 15, 2026

Domain-Independent Game Abstraction using Word Embedding Techniques

Many games of interest in the real world are often intractably large, thereby necessitating the use of game abstraction to shrink them in size, typically by many magnitudes. Over the last two decades, there have been significant advances in game abstraction; however, the domain-specific nature (usually poker) of much of the prior work prevents those techniques from being easily generalized to other settings without extensively analyzing the game at hand. In this paper, we propose a domain-independent approach to game abstraction, which applies word embedding techniques from the field of natural language processing. Treating each action as a word and gameplay data as a corpus, word vectors can be trained to represent each action as a real-valued vector, which can then be clustered to facilitate game abstraction. We also explore the use of foundational embedding models and show that action embeddings obtained this way can capture a surprising amount of information about the underlying game. Experimental results demonstrate that our proposed game abstraction technique is effective, although it does not outperform specialized algorithms tailored to specific games.

Juho Kim, Tuomas Sandholm

Computer Science Department Carnegie Mellon University · Computer Science Department, CMU · Strategic Machine, Inc. +2

Word Embeddings

Ask PDF

Sep 28, 2026

SAGE: Structured Strategic Reasoning for Efficient LLM Game Playing

A strong LLM strategic agent should reason prospectively over uncertain futures, adapt its strategy to opponents' behavioral tendencies, and continuously recalibrate its decision process from interaction experience. However, incorporating these sources in free-form reasoning could lead to unsupported strategic assumptions, inconsistent opponent estimates, and harmful interference from irrelevant historical interactions. To address these issues, we propose SAGE, a training-free inference-time framework that structures LLM strategic reasoning around three coordinated operations: anchor, adapt, and recalibrate. SAGE first anchors reasoning to an equilibrium policy that provides a strategically valid prior. It then conditions deviations from this anchor on a soft belief over opponent behavioral tendencies, enabling opponent-specific exploitation. Finally, SAGE distills strategically related interactions into counterfactual hypotheses about previously missing considerations, allowing past experience to recalibrate the model's reasoning. We evaluate SAGE on three repeated imperfect-information games: Leduc Hold'em, Liar's Dice, and Goofspiel, against various opponent types in each game. Compared with reasoning-intensive LLM agents, including Suspicion-Agent, ReTA, Agent-Pro, EMO, and Hypothetical Minds, SAGE achieves up to a 127.6% payoff improvement in Liar's Dice while reducing input and output token usage by up to 80% and 90%, respectively. In direct match-up play, it attains non-negative mean payoff against 5/10, 8/10, and 8/10 evaluated opponents in Leduc Hold'em, Liar's Dice, and Goofspiel, respectively, while using relatively fewer tokens. Code is available at https://github.com/chenzhwsysu57/SAGE.

Zhiwei Chen, Tianchun Wang, Zhongtao Rao +3

The Hong Kong University of Science and Technology (Guangzhou) · Johns Hopkins University · Microsoft +2

Game-Playing Agents Imperfect-Information Games

Ask PDF

Apr 28, 2026

StratFormer: Adaptive Opponent Modeling and Exploitation in Imperfect-Information Games

We present StratFormer, a transformer-based meta-agent that learns to simultaneously model and exploit opponents in imperfect-information games through a two-phase curriculum. The first phase trains an opponent modeling head to identify behavioral patterns from action histories while the agent plays a game-theoretic optimal (GTO) policy. The second phase progressively shifts the policy toward best-response (BR) exploitation, guided by a per-opponent regularization schedule tied to exploitability. Our architecture introduces dual-turn tokens -- feature vectors constructed at both agent and opponent decision points -- coupled with bucket-rate features that encode opponent tendencies across five strategic contexts. On Leduc Hold'em, a small poker variant with six cards and two betting rounds, we test against six opponent archetypes at two strength levels each, with exploitability ranging from 0.15 to 1.26 Big Blinds (BB) per hand. StratFormer achieves an average exploitation gain of +0.106 BB per hand over GTO, with peak gains of +0.821 against highly exploitable opponents, while maintaining near-equilibrium safety.

Andy Caen, Mark H. M. Winands, Dennis J. N. J. Soemers

Department of Advanced Computing Sciences, Maastricht University, Maastricht, The Netherlands

Opponent Modeling Game-Playing Agents Transformer Imperfect-Information Games Game Theory Multi-Agent Reasoning

Ask PDF

Abstract

Imperfect-Information Games

Explore similar work

Domain-Independent Game Abstraction using Word Embedding Techniques

SAGE: Structured Strategic Reasoning for Efficient LLM Game Playing

StratFormer: Adaptive Opponent Modeling and Exploitation in Imperfect-Information Games