---
title: Easy-to-Hard Generalization in Learning Systems
url: https://www.emergentmind.com/topics/easy-to-hard-generalization
type: topic
---

# Easy-to-Hard Generalization in Learning Systems

Easy-to-hard generalization denotes the ability of a learning system to acquire competence from easier instances, easier subtasks, or easier forms of supervision and then transfer that competence to harder instances that preserve the same underlying structure [2108.06011]. In recent work, the term spans at least three closely related settings: difficulty-controlled benchmarking, curriculum learning that orders data from easy to hard, and scalable supervision pipelines in which evaluators or self-improvement procedures trained on easier tasks are used to tackle harder reasoning problems [2603.27226; 2403.09472; 2502.01612]. The common research question is whether models learn the underlying rule or merely exploit superficial shortcuts that fail as difficulty increases [2108.06011]. The empirical record is mixed: some studies report strong transfer or gains from carefully designed curricula, whereas others find that difficulty-based ordering offers no robust advantage over random sampling and that transfer is often local in difficulty space rather than global [2401.06751; 2511.21692; 2603.27226].

## 1. Core definitions and problem formulations

One formalization of curriculum learning introduces a difficulty function
\[
f:\mathcal{X}\to \mathbb{N}_{>0},
\]
where larger \(f(x)\) denotes harder examples; a curriculum strategy then specifies which difficulty levels are sampled at each training phase [2603.27226]. In the canonical easy-to-hard variant, training begins with low difficulty and gradually expands to higher levels. In post-training studies of large language models, this ordering is compared directly against uniform random sampling while holding the total training budget fixed, so that the effect of ordering itself can be isolated [2603.27226].

A second formulation treats the topic as **cross-difficulty generalization**: train on examples of one difficulty and evaluate on another [2511.21692]. This includes both easy-to-hard generalization, in which training uses easier examples and evaluation uses harder examples, and hard-to-easy generalization, in which the direction is reversed [2511.21692]. In this setting, the central object is not the curriculum schedule but the train–test difficulty gap.

A third formulation makes difficulty explicitly learner-relative. The notion of **transitional problems** defines a problem \(p\) as transitional at level \(k\) if all weaker models fail on it and all models at or above level \(k\) succeed:
\[
\forall i \in [k-1]: \phi_p(M_i)=0 \quad \land \quad \forall i \in [k,n]: \phi_p(M_i)=1.
\]
This yields a model-specific easy-to-hard partition and a “level-up” curriculum that targets problems just beyond the current competence tier [2603.13761]. A related task-centric theory organizes reasoning problems into difficulty-indexed distributions \(p_1,\dots,p_L\), with easy-to-hard curricula moving progressively across this family rather than training on a fixed mixture [2602.10014].

These formulations differ in what “easy” and “hard” mean. Some work ties hardness to human ability to correctly label the data, using measures such as grade level, expert rating, Bloom-style cognitive skill, or reasoning-step count [2401.06751]. Other work argues that human proxies can be poor measures of what is difficult for LLMs, and instead derives difficulty from large-scale model behavior via Item Response Theory (IRT) or related systems [2511.21692; 2409.18433]. A recurring implication is that easy-to-hard generalization is not defined by a single universal difficulty axis.

## 2. Difficulty measurement and benchmark construction

Difficulty-controlled benchmarking is central to the literature because claims about easy-to-hard transfer depend strongly on how hardness is operationalized. Some benchmark families use transparent structural axes. In the dataset suite introduced specifically for studying generalization from easy training examples to harder test examples, Prefix Sums uses sequence length, Mazes uses maze size, and Chess Puzzles uses puzzle Elo difficulty rating [2108.06011]. In deductive-reasoning post-training, MathGAP-derived LinearDepth defines difficulty as the number of inference steps, PartWhole defines it as the number of axioms, and Knights-and-Knaves defines it as the number of characters, with the authors emphasizing that difficulty is structural rather than merely a surface proxy such as prompt length [2603.27226].

Other benchmark programs derive difficulty from performance statistics. A large-scale cross-difficulty study ranks items in ARC, GSM8K, MMLU-Pro, BBH, MATH, and MuSR using thousands of LLM responses and the Rasch / 1PL IRT model,
\[
P(r_{ij} \mid \theta_j, \beta_i) = \frac{1}{1 + e^{-(\theta_j - \beta_i)}},
\]
then partitions each dataset into 10 equal-sized bins from easiest to hardest [2511.21692]. Easy2Hard-Bench extends this idea into a standardized collection of six datasets with continuous numerical difficulty scores estimated from either IRT or Glicko-2, depending on domain [2409.18433].

| Benchmark family | Tasks | Difficulty definition |
|---|---|---|
| [2108.06011] | Prefix Sums, Mazes, Chess Puzzles | Longer strings, larger mazes, higher Elo difficulty rating |
| [2603.27226] | LinearDepth, PartWhole, Knights-and-Knaves | Inference steps, number of axioms, number of characters |
| [2511.21692] | ARC, GSM8K, MMLU-Pro, BBH, MATH, MuSR | 10 equal-sized IRT bins from easiest to hardest |
| [2409.18433] | E2H-AMC, E2H-Codeforces, E2H-Lichess, E2H-GSM8K, E2H-ARC, E2H-Winogrande | Numerical difficulty scores from IRT or Glicko-2 |

A major theme in this line of work is the divergence between human-oriented and model-oriented hardness. The IRT-based study reports that grade level, question length, answer length, reasoning-step count, Bloom’s taxonomy, Depth of Knowledge, and expert ratings correlate weakly or inconsistently with model-based difficulty; examples include a positive correlation up to about \(0.56\) for MATH answer length and about \(0.49\) for GSM8K reasoning steps, but many other correlations are near zero or negative [2511.21692]. Easy2Hard-Bench likewise argues for continuous, unified, performance-derived scores rather than coarse labels or domain-specific heuristics [2409.18433]. This suggests that a central technical issue in easy-to-hard research is not only transfer, but calibration of the difficulty axis itself.

## 3. Training regimes and algorithmic patterns

Static easy-to-hard curricula remain the most direct operationalization. In deductive-reasoning post-training, five sampling strategies are compared: Standard, SingleDiffInc, SingleDiffDec, UpToDiff, and DownToDiff, under fixed epoch and sample budgets across SFT, GRPO, and PPO [2603.27226]. This setting isolates sequencing from confounds such as adaptive pacing or varying compute. Closely related work on mathematical reasoning evaluates Forward Curriculum Learning, Reverse Curriculum Learning, and grouped variants under identical offline fine-tuning settings, explicitly reframing curriculum quality as a question about data ordering rather than dataset content [2510.19099].

A more competence-aware family of methods constructs the curriculum relative to the learner. Transitional-problem curricula partition the data into subsets \(\mathcal{D}_\tau\) defined by the level at which problems flip from unsolved to solved across a monotonic model series, then train in Level-Up, Level-Down, or Random order [2603.13761]. The distinctive claim is that the useful frontier is not globally easiest-first, but the set of problems just beyond the model’s current ability.

Several lines of work replace direct easy-to-hard imitation with auxiliary supervision. In scalable alignment for mathematical reasoning, process-supervised reward models are trained only on easy problems and then used to evaluate candidate solutions to hard problems, enabling reranking or RL without human supervision on the hard subset [2403.09472]. In iterative self-improvement, the model generates pseudo-labels on slightly harder inputs and is retrained on the growing dataset; at round \(r\),
\[
\mathcal{D}_r = \{(x_i, M_{r-1}(x_i))\}_{i=1}^{N_r},
\]
with \(d_{r-1}\le \text{Difficulty}(x_i)\le d_r\), so that each round moves only slightly beyond previous competence [2502.01612].

Curriculum reinforcement learning makes the schedule part of the RL objective. E2H Reasoner decomposes a hard task into progressively more difficult subtasks and samples from them according to schedules such as a cosine fade from easy to hard or a Gaussian schedule that more aggressively fades out easy tasks [2506.06632]. This differs from simple easy-first staging because the method explicitly treats prolonged exposure to easy tasks as a source of overfitting.

Beyond generic reasoning, easy-to-hard structures are also embedded into task-specific architectures. FE2H for multi-hop question answering uses a two-stage selector that retrieves one supporting document and then conditions on it to retrieve the second, and a two-stage reader trained first on SQuAD and then on HotpotQA [2205.11729]. CAFE for panoptic scene graph generation aligns predicate difficulty with feature complexity, moving from bbox features to mask features to boundary features [2407.09191]. MetaLint trains on easy Ruff-detectable idioms and evaluates on harder PEP-inspired idioms outside the linter’s direct coverage [2507.11687].

## 4. Empirical findings: strong transfer, local transfer, and negative results

A substantial body of evidence reports surprisingly strong easy-to-hard transfer under some conditions. In question answering with pretrained LMs, easy supervision often recovers **70%–100% of the supervision gap** on hard test data, and models trained on easy data can perform nearly as well as oracle models finetuned on hard data across ARC, MMLU, and StrategyQA, using in-context learning, linear classifier heads, and QLoRA [2401.06751]. That result is especially salient because it is framed by the scalable oversight problem: easy data may be cheaper and less noisy than hard data, yet still competitive for hard-test accuracy [2401.06751].

Evaluator-guided supervision yields a different positive result. In mathematical reasoning on MATH, direct easy-to-hard SFT is weaker than full SFT, but reward models trained on easy problems generalize better than generators trained on easy problems, and can then improve hard-task performance through reranking or RL [2403.09472]. A process-supervised 7b RL model reaches **34.0%** on MATH500 and a 34b reranking system reaches **52.5%** on MATH500 despite using human supervision only on easy problems [2403.09472]. This is a strong form of easy-to-hard generalization because the bridge from easy supervision to hard performance is evaluation rather than generation.

Iterative self-improvement provides another positive regime. Standard transformers trained on easier examples and then iteratively retrained on self-generated solutions can generalize from **1–16 digits** to **over 100 digits** in reverse addition, from **1–10** to **over 120 digits** in copy and reverse, maintain **>98% accuracy up to length 70** in forward addition with filtering, extend multiplication from **\(5\times 5\)** to **\(9\times 9\)** near-perfectly, and generalize maze solving to **30-hop** paths when filtering is used [2502.01612]. Controlled weak-to-strong curricula and filtering are essential in these results; without filtering, several tasks collapse [2502.01612].

Learner-specific curricula also produce gains. In chess and mathematics, Level-Up training on transitional problems consistently outperforms random ordering, hard-to-easy ordering, and curricula based on external heuristics such as length, number of steps, legal moves, or human ratings [2603.13761]. The claim here is not that any easy-first schedule works, but that the right easy-to-hard order is the one defined relative to the learner’s competence.

Against these positive results stand several prominent negative or strongly qualified findings. In deductive-reasoning post-training on synthetic arithmetic and logical benchmarks, difficulty-based sequencing shows **no robust advantage** over standard uniform random sampling in either accuracy or response length, and this remains true across multiple model families, curriculum schedules, SFT, and RL [2603.27226]. The paper’s conclusion is not that post-training fails, but that the gains from post-training itself are much larger than any gains from curriculum order [2603.27226].

Cross-difficulty transfer is also often sharply local. Fine-tuning on a single IRT difficulty bin yields the best transfer to nearby bins, and performance drops as the train–test difficulty gap widens; with sufficiently large gaps, performance can even fall below the zero-shot baseline [2511.21692]. Easy2Hard-Bench reports a similar **generalization ridge** along the diagonal where training difficulty is close to evaluation difficulty, and finds that easy-to-hard transfer is not uniformly strong [2409.18433]. These studies reject the idea that easy data or hard data alone can serve as a universal substitute across the full difficulty range.

The literature on curriculum direction further complicates the picture. A unified study of mathematical reasoning concludes that **no curriculum strategy dominates universally**: the relative value of forward versus reverse curriculum depends jointly on the difficulty metric, the model’s capability, and task complexity [2510.19099]. In another line of work, meta-learning with learned reweighting finds that using hard validation data performs consistently better than using easy validation data and improves classifier margins, implying that in some settings the hard subset is the more informative supervisory target [2403.12236]. Taken together, these results suggest that easy-to-hard generalization is real but conditional, and that its success depends on how supervision, difficulty, and optimization are coupled.

## 5. Mechanistic explanations and theoretical accounts

A recurrent explanation for positive results is that **evaluation is easier than generation**. In scalable alignment for mathematical reasoning, direct easy-to-hard generation is weak, but easy-to-hard evaluation is stronger: process reward models trained on easier tasks can score candidate hard-task solutions well enough to support reranking and RL [2403.09472]. This reframes the problem from “generate hard answers from easy demonstrations” to “learn an evaluator on easy tasks that still recognizes good hard-task reasoning.”

Another explanation centers on competence-aware frontiers. Transitional-problem curricula assume that the most useful training examples are those that sit just beyond the current learner’s ability, rather than the globally easiest examples [2603.13761]. The educational intuition is explicit: tasks near the next competence tier are maximally useful for “leveling up.” This suggests that the practical value of easy-to-hard ordering depends less on abstract difficulty and more on alignment between task difficulty and current policy quality.

Theoretical analyses of self-improvement formalize a positive feedback loop. In a task-centric theory, each iteration fine-tunes on a reward-filtered distribution, and better models accept more samples per iteration, increasing the effective training set and improving subsequent updates [2602.10014]. The framework defines a global acceptance rate
\[
Z_{p}(\theta)=\mathbb E_{q\sim p}\big[\alpha(\theta,q)\big],
\]
and shows that this acceptance-mediated data expansion supports sustained improvement but also explains eventual saturation [2602.10014]. In the binary-reward case, the theory derives a recursive lower bound
\[
V_{p_0}(\hat\theta_{t+1}) \ge F\!\left(V_{p_0}(\hat\theta_t)\right),
\]
making the feedback loop explicit [2602.10014].

Curriculum RL adds a complementary API-based account. E2H Reasoner proves convergence guarantees and finite-sample complexity bounds under a decomposition of a hard task into progressively harder subtasks, and argues that when tasks are appropriately decomposed and conditioned, learning through curriculum stages requires fewer total samples than direct learning on the hard distribution [2506.06632]. A key practical claim is that easy tasks are useful only initially; they must be faded out through appropriate scheduling to avoid overfitting [2506.06632].

Negative results motivate different mechanisms. The deductive-reasoning study argues that if models mostly learn surface-level patterns rather than abstract compositional rules, then example ordering should matter little; the invariance of response length across curricula is presented as evidence that curriculum order does not alter the model’s effective reasoning depth [2603.27226]. The large-scale IRT study similarly argues that easy training data often lacks the coverage needed for harder decision boundaries, producing transfer that is strongest only near the training difficulty band [2511.21692]. A plausible synthesis is that ordering helps when the curriculum is competence-aware or verifier-supported, but not when the underlying training signal fails to induce rule learning.

## 6. Extensions, applications, and unresolved issues

Easy-to-hard generalization has been instantiated in multiple application domains beyond generic reasoning benchmarks. In multi-hop question answering, FE2H uses easy-to-hard progression in both document selection and reader transfer, reporting **71.89** Answer EM, **84.44** Answer F1, **64.98** Supporting EM, **89.14** Supporting F1, **50.04** Joint EM, and **76.54** Joint F1 on HotpotQA distractor, while the two-stage selector reaches **96.32** EM and **98.02** F1 on document selection [2205.11729]. Here, “easy” refers both to the first retrieval hop and to single-hop QA pretraining before multi-hop transfer.

In panoptic scene graph generation, CAFE turns easy-to-hard progression into a curriculum over feature complexity. Easy predicates are learned with bbox features, medium predicates add mask features, and hard predicates add boundary features; the curriculum and distillation strategy improve both robust and zero-shot PSG, with Transformer + CAFE reaching **46.1** average in zero-shot PredCls compared with **37.6** for the Transformer baseline [2407.09191]. This is an instance in which the easy-to-hard axis applies to predicates and representational richness rather than to sample count or sequence length.

In machine-generated text detection, an easy-to-hard supervision framework trains an easier supervisor on longer texts, where class boundaries are clearer, and uses structural coupling to improve the harder target detector on shorter or more ambiguous texts [2511.00988]. Across **312 cross-LLM settings**, the enhancement beats the baseline in **87%** of sentence-level cases and **85%** of paragraph-level cases [2511.00988]. In code quality analysis, MetaLint trains on Ruff-detectable idioms and transfers to unseen PEP-inspired idioms, reaching **70.37%** detection F-score, **70.43%** detection recall, and **26.73%** localization on the hard benchmark [2507.11687]. These systems treat “easy” as automatically verifiable or structurally clearer supervision.

The topic also extends beyond language models. In CNN training on CIFAR-10 and ImageNet-1k, easy examples are defined by low early-training loss averaged across runs, hard examples by high early-training loss; easy examples are visually similar, hard examples visually diverse, and both sets are largely shared across architectures [1911.10739]. Hard examples contribute more individually to generalization, but removing too many easy examples still harms accuracy, and a stepwise easiness-based removal strategy yields about **1.1% accuracy** gain over random removal at **0.3 ablation ratio** on ImageNet-1k [1911.10739]. This complicates any simplistic view that easy examples are merely redundant.

Several unresolved issues recur across the literature. Difficulty may be structural, human-defined, model-defined, or learner-relative; these notions do not reliably coincide [2511.21692; 2409.18433]. Many conclusions come from synthetic or highly controlled tasks, medium-sized models, fixed training budgets, or narrow domain coverage, and several papers explicitly do not study adaptive curricula, convergence speed under alternate stopping criteria, or broader multilingual settings [2603.27226; 2409.18433]. The resulting picture is therefore not a universal doctrine of “start easy, end hard,” but a heterogeneous research area in which transfer can be strong, weak, or absent depending on how difficulty is measured, how supervision is delivered, and whether the training signal targets rule learning, calibration, or search.

Source: https://www.emergentmind.com/topics/easy-to-hard-generalization