Neural-symbolic integration

Overview

Neural networks learn from data but, by the account of the position paper, are brittle outside their training distribution, hard to explain and data-hungry 2012.05876 §2. Symbolic systems reason precisely over facts and rules but learn poorly from raw perception 2012.05876 §2 2012.05876 §4. Neural-symbolic integration combines the two so that a system gets the strengths of both 2012.05876 §2. The survey frames the same goal as a question of representation: which logic is encoded, and how learning, reasoning and extraction work on that encoding 1905.06088 §1 1905.06088 §2.

The invariant that the position paper asks for is a principled two-way link between the two representations. Knowledge can be compiled into a network, and symbolic descriptions can be extracted from a trained one 2012.05876 §2. Not every variant has both directions. Binder trains nothing and uses the language model only through prompts and calls inside a symbolic program 2210.02875 §2.1 2210.02875 §2.2. The binding-problem paper takes a different position: object-level structure should be learned inside the network, not handed to a separate symbolic system 2012.05208 §7.

The position paper calls the choice of representation the first design decision: whether knowledge is held as discrete named symbols (localist) or as continuous vectors (distributed) 2012.05876 §3. The survey uses the same kind of split to organise its methods, by representation, learning, reasoning and explainability 1905.06088 §3–§6.

Variants

The position paper's own taxonomy is Kautz's six types, which the authors re-read through the localist/distributed distinction and extend in places 2012.05876 §3. The variants below group those types, and the survey's methods, by how the symbolic and neural parts are joined.

Hybrid network and symbolic solver

A network handles perception and a separate symbolic reasoner handles inference. The two exchange data through an interface, so the solver can be exact and checkable 2012.05876 §4. The position paper names AlphaGo, which pairs a network with Monte Carlo tree search, as a loosely coupled example (Type 2), and the neuro-symbolic concept learner and deepProbLog as examples where the two parts work on complementary tasks (Type 3) 2012.05876 §3. The survey's vertical hybrids follow the same pattern: a logic layer sits over a perception network, as in DeepProbLog, where network outputs act as neural predicates 1905.06088 §4.3. The authors see this pattern as the likely practical choice 2012.05876 §4. Card: Neurosymbolic AI: The 3rd Wave.

Neural calls inside a symbolic program

A language model writes a program in a symbolic language such as SQL or Python. Where the language has no operator for a sub-task, the program holds a call back to the same model, asking a small question about a few cells or columns. The program is executed, and several sampled programs vote on the answer 2210.02875 §2.1 2210.02875 §2.2. The method needs only a few annotated program exemplars and no training 2210.02875 §3.1. Card: Binding Language Models in Symbolic Languages.

Language model writes a program or problem for a solver

The model does not produce the answer. It writes a formal artifact, and a fixed engine computes the result. In PAL the artifact is a short Python program with each step written as a comment, and the interpreter's output is the answer. The method uses a few worked examples and no training 2211.10435 §3. In LLM+P the model writes a PDDL problem file from a task description, guided by one example pair. The classical planner Fast Downward then finds a plan with its optimal setting, and the domain file is written in advance for each domain 2304.11477 §3.2 2304.11477 §3.3. In both, the answer is only as good as the artifact, since the interpreter or planner is exact 2211.10435 §6 2304.11477 §5.3. Cards: PAL: Program-aided Language Models, LLM+P: Empowering Large Language Models with Optimal Planning Proficiency.

Logic-LM applies the same split to four formal languages: logic programs, first-order logic, constraint satisfaction, and analytical puzzles posed as constraints. Each has its own solver, from Pyke and Prover9 to python-constraint and Z3 2305.12295 §3.1 2305.12295 §3.2. The paper's argument is that the model only has to represent the problem, not solve it 2305.12295 §3, and that the answer is faithful only as far as the formulation is correct 2305.12295 §3. When the solver reports an execution error, the program, the message and examples of common errors go back to the model for a revision, up to a limit 2305.12295 §3.3. Card: Logic-LM: Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning.

Select-then-infer loop without a solver

Selection-Inference has no solver at all. The language model is the only engine, and the neural and symbolic parts are split into two calls per step 2205.09712 §4. The selection call picks facts from the context. It scores candidate facts rather than generating them, because the authors found that free generation invented facts 2205.09712 §4.1. The inference call then sees only the selected facts, not the question, and writes one new fact, which is appended to the context 2205.09712 §4.2. The number of steps is fixed, and the paper leaves a learned stop to future work 2205.09712 §4 2205.09712 §7. No step is checked by an exact engine, so the trace is readable but not verified. Card: Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning.

Knowledge compiled into the network

Symbolic knowledge is translated into the training data or into the initial architecture and weights of a localist network. The position paper extends this type from if-then rules to general symbolic knowledge 2012.05876 §3. In the compiler view, the knowledge either sets initial weights or adds a knowledge-based penalty to the loss 2012.05876 §2. The survey lists early examples such as KBANN, which stacks one perceptron per implication, and CILP 1905.06088 §3.1.1. It also describes horizontal hybrid learning, in which rules are combined with data during training, and self-transfer, where rules extracted from a teacher network regularise a student 1905.06088 §4.2. The informed-machine-learning survey gives the same kind of example: rules encoded in a network with KBANN, and a knowledge term added to the usual loss 1903.12394 §5.5 1903.12394 §5.1. It reports that logic rules often enter the hypothesis set, while simulation results most often enter the training data 1903.12394 §3.2.3. Cards: Neurosymbolic AI: The 3rd Wave, Neural-Symbolic Computing: An Effective Methodology, Informed Machine Learning: A Taxonomy and Survey.

Logic as differentiable constraints

First-order formulas become differentiable constraints with truth values in [0, 1], so they shape an embedding during training and the trained network can be queried for whether a statement holds 2012.05876 §2 2012.05876 §3. Logic Tensor Networks are the main example, and tensor product representations are named as a related approach 2012.05876 §3. The survey places LTNs among its tensorised first-order methods, as an extension of Neural Tensor Networks to whole formulas 1905.06088 §3.2.2. In this type a rule acts as a soft constraint, or regulariser, on the loss 2012.05876 §3. Cards: Neurosymbolic AI: The 3rd Wave, Logic Tensor Networks.

The LTN paper turns the whole knowledge base into one satisfiability score, which training pushes up by gradient descent. The connectives and quantifiers use a stable product configuration of fuzzy operators 2012.13635 §2.4 2012.13635 §3.2. In its digit-addition setup the network sees only the sum of two images. A rule allows only the digit pairs that add up to that sum, so the digits are learned from the sums alone 2012.13635 §4.4. The same formalism covers classification, regression and clustering, but only the digit-addition results have a baseline 2012.13635 §4.

Reasoning inside a neural engine

The aim is symbolic reasoning, ideally combinatorial reasoning, carried out inside the network itself (Type 6) 2012.05876 §3. The position paper says that no complete Type 6 system for combinatorial reasoning existed at the time of writing, and lists goal-directed commonsense and combinatorial reasoning as an open challenge 2012.05876 §3 2012.05876 §6. The survey describes differentiable forms of this idea: a neural theorem prover that extends backward chaining, and neural logic machines that take facts as input tensors and output new facts 1905.06088 §4.1. It also states that exact maximum satisfiability is NP-hard, without a citation, and that a neural approximation gives feedforward inference instead 1905.06088 §5.2. Card: Neurosymbolic AI: The 3rd Wave.

Rule extraction from a trained network

The decompiler direction: a symbolic description is pulled out of a trained network 2012.05876 §2. The position paper uses it to support explanations that show a proof trace, which a raw activation pattern cannot 2012.05876 §5. Extraction from very large networks, with soundness guarantees, is listed as the first open challenge 2012.05876 §6. The survey judges extraction methods on accuracy, fidelity, consistency and comprehensibility, and reports that global combinatorial search does not scale to current networks 1905.06088 §6.1. It also reports that clauses pulled from the first layer of a deep belief network work well, while higher layers lose accuracy when treated the same way 1905.06088 §6.1.

A KAN gives another route to a formula. Each edge is trained with a sparsity penalty, the network is pruned, and each surviving edge can then be fixed to a named function such as a sine or an exponential by a grid search and a linear fit 2404.19756 §2.5.1. The authors apply this by hand to small examples, for instance a [2,5,1] network that is reduced to [2,1,1] and then fixed to symbols 2404.19756 §2.5.2, and they report recovered formulas for science problems 2404.19756 §4. The fit is per edge and manual, and the interpretability claims rest on small examples [arxiv:2404.19756 §2.5.1] 2404.19756 §6. Card: KAN: Kolmogorov-Arnold Networks.

Symbolic beliefs about other agents

An embodied agent that must infer what people believe, want or intend can use either of two paradigms 2601.02378 §4. In Prompting, a language model reasons about mental states through prompt design or context engineering, with no fine-tuning 2601.02378 §4.1. SymbolicToM, which tracks symbolic beliefs, is one method listed in this group [arxiv:2601.02378 tab. 4]. In Model-Based Inference, an explicit probabilistic model of the other agent is inverted to infer its state from its observed actions 2601.02378 §4.2. The review's central trade-off is interpretability against flexibility 2601.02378 §3. It names a combination of symbolic beliefs and distributed activation vectors, by decoding symbolic belief units from LLM hidden states, as the most interesting recent direction 2601.02378 §3. It also names neural-symbolic integration as the main direction in two forms: neural generation with symbolic verification, and symbolic guidance with neural fine-tuning 2601.02378 §6.1. Card: Modeling the Mental World for Embodied AI: A Comprehensive Review.

Object representations learned end to end

The binding-problem paper argues that networks cannot dynamically bind distributed information into object-like representations, and that this limits systematic generalisation 2012.05208 §2.3. It splits the problem into representation, segregation and composition [arxiv:2012.05208 §3] 2012.05208 §4 2012.05208 §5. It judges slot-based representations the simplest mechanism, while augmented distributed codes and tensor product representations are harder to build but support more sophisticated use 2012.05208 §3.4. It argues that object learning should be largely unsupervised, fully differentiable and learned end to end, rather than handed to a separate symbolic system 2012.05208 §4.4 2012.05208 §7. The paper has no experiments, so these are design arguments. Card: On the Binding Problem in Artificial Neural Networks.

Neural-symbolic search over knowledge graphs

Knowledge-graph reasoning mixes the two parts in a different way. For a single missing fact, the survey lists symbolic rules, embeddings, and neural-symbolic variants such as neural rule learning, which learns chains of relations with a recurrent network 2412.10390 §3.1 2412.10390 §3.2.2. For logical queries, a link predictor gives each triple a fuzzy truth value in [0, 1], and a solver searches for the entity assignments that satisfy the query. Beam search covers acyclic query graphs, and a continuous relaxation in embedding space allows gradient-based search 2412.10390 §4.3. The survey reports no numbers for these methods. Card: Neural-Symbolic Reasoning over Knowledge Graphs: A Survey from a Query Perspective.

Learned mapping between two embodiments

A robot learns from a person's video. The reward and policy come from adversarial inverse reinforcement learning, fed with 3D human keypoints and object positions 2412.11360 §IV.A. A learned model then maps the human arm pose to initial joint angles on a cobot, using a fixed one-to-one joint correspondence, and an inverse-kinematics step refines the angles on a learned forward-kinematics model 2412.11360 §IV.B fig. 3. The authors call this mapping neuro-symbolic, but its symbolic part is the joint table, which the text describes without equations 2412.11360 §IV.B. Against RRT and RRT-connect, the authors report shorter and smoother motion on two cobots 2412.11360 tab. 5 2412.11360 tab. 6. Card: Visual IRL for Human-Like Robotic Manipulation.

When to use which

The position papers and surveys give design arguments, not measured comparisons. Binder, PAL, LLM+P, Logic-LM, Selection-Inference and KAN report measured numbers, but only their authors' own. The rows below follow the stated constraints. Test each one against your own task before relying on it.

Implementation sketch

No code is released with the position papers or the surveys. The patterns below are our own sketches of the papers' descriptions, with the names chosen by us. The Binder pattern is the hybrid loop with the facts produced on demand, as calls the program makes back to the model 2210.02875 §2.1. The generate-then-solve pattern covers PAL, LLM+P and Logic-LM. Logic-LM adds a repair loop, in which a failing program is rewritten from the solver's error message 2305.12295 §3.3.

Hybrid loop: the network perceives, the solver reasons.

# Hybrid prediction: network perceives, solver reasons  [arxiv:2012.05876 §4]
def hybrid_predict(x, rules, solver):
    # Map the raw input to symbol scores with the trained network  [§4]
    symbols = network.perceive(x)
    # Convert scores to facts through the neurosymbolic interface  [§4]
    facts = to_facts(symbols)
    # Exact reasoning over expert rules; the solver stays symbolic  [§6]
    return solver.query(facts, rules)

Soft-constraint training: rules shape the loss.

# Training with rules as soft constraints  [arxiv:2012.05876 §2] [§6]
def train_step(batch, rules, lam):
    # Standard task loss on the labelled data  [§6]
    loss = task_loss(network(batch.x), batch.y)
    # Penalise rule violations; the paper gives no weighting scheme  [§6]
    loss = loss + lam * rule_violation(network, rules, batch.x)
    return loss

Generate-then-solve: the model writes an artifact and an exact engine answers.

# Generate a formal artifact, then let an exact engine answer
# [arxiv:2211.10435 §3] [arxiv:2304.11477 §3.3]
def generate_then_solve(task, examples, engine):
    # The model writes a program or problem file, not the answer  [§3]
    artifact = llm_generate(build_prompt(examples, task))
    # Reject artifacts the engine cannot parse; malformed problem files
    # are the main failure LLM+P reports  [arxiv:2304.11477 §5.3]
    if not engine.parses(artifact):
        return None
    # The engine's result is the answer, not the model's own sums  [§6]
    return engine.run(artifact)

For the Binder and PAL patterns, the model writes code that is then executed. Run it in a sandbox with a time limit, and treat it as untrusted.

Contradictions

Disagreements between papers, each side cited. Not resolved here.

• Where the symbols belong. The position paper calls a hybrid network and solver the likely practical choice 2012.05876 §4. The binding-problem paper argues that object learning should be integrated and learned end to end, rather than handed to a separate symbolic system 2012.05208 §7.
• Ranking under another evaluator. Binder reports 64.6 on WikiTQ test against 63.3 for OmniTab 2210.02875 §3.2. Under the official evaluator, the paper reports 61.9 for Binder against 62.9 for OmniTab 2210.02875 app. C.2.
• Stepwise reasoning against chain-of-thought. Selection-Inference reports that its loop beats chain-of-thought with a 7B Gopher model on 10 bAbI and ProofWriter tasks 2205.09712 §5. Logic-LM reports that GPT-4 with chain-of-thought scores higher than its own method on PrOntoQA, 98.79 against 83.20 2305.12295 tab. 2. The models, tasks and prompts differ, so the two results do not measure the same setting.
• Type 6 wording. The position paper says early neural-symbolic work achieved symbolic reasoning inside a neural engine 2012.05876 §3, and also that no complete Type 6 system for combinatorial reasoning existed 2012.05876 §3. The two differ in scope.

Limits of the evidence

Binder, PAL, LLM+P, Logic-LM, Selection-Inference KAN and Logic Tensor Networks report measured numbers, all author-reported. Binder's come from one model, Codex code-davinci-002 2210.02875 §3.1, which is retired, and the official evaluator changes the ranking against OmniTab 2210.02875 app. C.2. PAL's come from the same Codex model 2211.10435 §4, and the paper's ablations show the gain depends on the interpreter 2211.10435 §6. LLM+P's come from GPT-4 with one example per domain 2304.11477 §3.2 2304.11477 §5.1. Its optimal numbers hold only where the optimal planner finishes in 200 seconds; timeouts fall back to a sub-optimal setting shown in parentheses 2304.11477 §5.2. KAN's come from small function-fitting, PDE and science tasks 2404.19756 §3. Its accuracy and scaling results assume the true network shape is known 2404.19756 §3.1, and for one toy shape the measured scaling exponent is about G^-3 against the predicted G^-4 2404.19756 §2.4. Its formula recovery needs a human to name each candidate function 2404.19756 §2.5.1. Logic-LM's numbers come from GPT-4, GPT-3.5 and text-davinci-003 on five test sets 2305.12295 §4. Its abstract averages, 39.2% over standard prompting and 18.4% over chain-of-thought, are attributed to GPT-3.5 in the text [arxiv:2305.12295 §4.1]. The card recomputes them as means over the three base models, and some §4.1 figures do not follow from the paper's own table; see the card. The released code also differs from the paper in its fallback when no answer is produced, so the paper's fallback should not be assumed 2305.12295 §4.1. Selection-Inference's headline number comes from a 7B Gopher model, 5-shot, on 10 tasks 2205.09712 §5. Its prompt examples are not specified, and its baseline figures in §6 differ from those in §5, so the §5 and §6 numbers should not be mixed 2205.09712 §5 2205.09712 §6. The fine-tuned variant is trained on single steps only 2205.09712 §6. The six papers that are position papers or surveys run no experiments. Their claims about efficiency, modularity, explanation and fidelity are arguments 2012.05876 §2 2012.05208 §7 1905.06088 §6 2412.10390 §1 1903.12394 §3.1.2 2601.02378 §1. The position paper's fidelity claim about LIME-style methods cites an earlier measurement 2012.05876 §5. Its statement that networks can represent propositional and some non-classical logics, but only part of first-order logic, is the basis of its first challenge. It should be checked against the cited work before it is relied on 2012.05876 §2 2012.05876 §6. Logic Tensor Networks' MNIST numbers are the mean of the 10 best of 15 runs, and the optimiser stalls at initialisation in about one run in five 2012.13635 §4.4. Its other experiments are shown as plots or described in words 2012.13635 §4.5 2012.13635 §4.7. The released code differs from the text in several example settings, such as the regression learning rate and the smokes embedding size; see the card. Visual IRL's motion results are measured against RRT and RRT-connect only, and its learned-behaviour accuracy is given in the text without a table 2412.11360 §V.D. Its object detector was trained on 3000 frames and tested on 500, all from one lab setup 2412.11360 §V.D. Its AIRL and TRPO settings are not stated 2412.11360 §IV.A. No official code was found for it.

Papers

• Neurosymbolic AI: The 3rd Wave: a position paper that maps neural-symbolic work onto Kautz's taxonomy, argues that localist versus distributed representation is the central design choice, and lists three open challenges. It has no experiments.
• Neural-symbolic computing survey: a survey that organises neural-symbolic methods by representation, learning, reasoning and explainability. It gives no new results.
• Binding Language Models in Symbolic Languages: Binder, a training-free method in which a language model writes SQL or Python programs with calls back to the model, and votes over sampled programs. It reports author-reported table-QA results.
• On the Binding Problem in Artificial Neural Networks: a position survey that splits the binding problem into representation, segregation and composition, and argues for learned object-level structure. It has no experiments.
• KAN: Kolmogorov-Arnold Networks: an architecture with learnable spline functions on edges, which can be pruned and fitted to named functions to give formulas. It reports author-reported results on small problems.
• Neural-Symbolic Reasoning over Knowledge Graphs: A Survey from a Query Perspective: a survey of knowledge-graph reasoning by query type, grouped into symbolic, neural and neural-symbolic families. It gives no numbers.
• PAL: Program-aided Language Models: few-shot prompting in which the model writes Python with comments and an interpreter gives the answer, with majority voting over samples. It reports author-reported gains over chain-of-thought with Codex on math and symbolic tasks.
• LLM+P: Empowering Large Language Models with Optimal Planning Proficiency: GPT-4 writes a PDDL problem file from one example, and an optimal classical planner solves it. It reports author-reported planning success on seven domains.
• Informed Machine Learning: A Taxonomy and Survey of Integrating Knowledge into Learning Systems: a survey with a three-axis taxonomy of how prior knowledge enters learning, covering logic rules, equations and invariances. It has no experiments of its own.
• Logic-LM: Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning: an LLM translates a logic problem into one of four formal languages, a deterministic solver runs it, and solver errors trigger revisions. It reports author-reported accuracy on five logical-reasoning datasets.
• Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning: a frozen LLM loop that selects facts from the context and derives one new fact per step, with no training and no solver. It reports author-reported accuracy on 10 logic tasks with a 7B model.
• Modeling the Mental World for Embodied AI: A Comprehensive Review: a survey that separates physical from mental world models, compares prompting and model-based theory-of-mind methods, and catalogues benchmarks. It runs no new experiments and quotes numbers from cited works.
• Visual IRL for Human-Like Robotic Manipulation: adversarial inverse RL from human keypoints, with a learned mapping to cobot joints and IK refinement. It reports author-reported results on two cobot tasks against RRT variants.
• Logic Tensor Networks: first-order logic over real-valued tensors, where a knowledge base's satisfiability is maximised during training. It reports author-reported MNIST addition results against DeepProbLog, and qualitative results elsewhere.