Scaling Laws & Emergent Abilities

Empirical power laws that predict how loss and capability improve with parameters, data, and compute, and how to choose compute-optimal training runs.

published · difficulty 3/5 · 18 min read

Reading map and next steps

Intuition

Build the mental picture first so the rest of the page has something to attach to.

PredictName the object in plain language, then predict what should change.Leave with one reusable mental picture before notation appears.

If you train a family of models the same way (same architecture class, data pipeline, optimizer recipe), then "bigger model" and "more data" usually give you predictably better loss.

Scaling laws turn that predictability into a planning tool. Instead of guessing, you can fit a curve from small runs, then forecast how far a larger run will go, and how to spend a fixed compute budget:

Should we buy more parameters or more tokens?
If we only get one big run, what is the compute-optimal choice?
Why do some task behaviors look like they "appear suddenly"?

The key mindset is: scaling laws are not a proof about intelligence. They are an empirical control system for allocating compute.

Section prompt

Copy-only prompts — each action copies a page-grounded prompt to your clipboard. Nothing is sent by this site.

Math

Translate the story into symbols, assumptions, and a derivation you can inspect.

InspectTrack the same object through the notation and check each symbol.Leave with the invariant the equations preserve.

Equation 1Inspect equation objectL(N, D) \approx L_infinity + a N^{-\alpha} + b D^{-\beta},Equation 2Inspect equation objectC \propto N\cdot D.Equation 3Inspect equation objectD \propto N.

Power-law loss scaling

A common empirical form for test loss $L$ is:

L(N, D) \approx L_\infty + a\,N^{-\alpha} + b\,D^{-\beta},

where:

$N$ is parameter count,
$D$ is number of training tokens (or examples),
$\alpha,\beta>0$ are exponents you fit from data,
$L_\infty$ is the irreducible loss floor for the dataset/model class.

On a log-log plot, $N^{-\alpha}$ and $D^{-\beta}$ look like straight lines. That's why power laws are useful: they extrapolate smoothly.

Compute-optimal allocation (Chinchilla-style rule of thumb)

Very roughly, training compute scales like:

C \propto N\cdot D.

Under this constraint, minimizing the loss above typically yields a near-linear rule:

D \propto N.

Interpretation: for a fixed compute budget, don't overscale parameters without scaling data too, or you spend compute learning the same patterns repeatedly.

"Emergent abilities" as sharp transitions

When you measure a capability with a thresholded metric ("accuracy above 50%", "passes a benchmark"), smooth curves in loss can turn into sharp-looking transitions. The underlying performance can still be continuous.

Section prompt

Copy-only prompts — each action copies a page-grounded prompt to your clipboard. Nothing is sent by this site.

Code

Keep the implementation aligned with the notation so the algorithm is legible.

TraceMatch variables to symbols before reading the implementation.Leave with a runnable witness for the math.

Code witness 1import numpy as np # A toy scaling law: L(N,D) = L_inf + a N^-alpha + b D^-beta L_inf, a, b =...python

import numpy as np

# A toy scaling law: L(N,D) = L_inf + a N^-alpha + b D^-beta
L_inf, a, b = 1.5, 1.0, 1.0
alpha, beta = 0.07, 0.095

def loss(N, D):
    return L_inf + a * N**(-alpha) + b * D**(-beta)

C = 1e12  # compute budget in arbitrary units ~ N*D
Ns = np.logspace(7, 10, 40)  # 10M..10B params

best = None
for N in Ns:
    D = C / N
    L = loss(N, D)
    if best is None or L < best[0]:
        best = (L, N, D)

L, N, D = best
print("best loss:", round(float(L), 4))
print("N:", f"{N:.2e}", "D:", f"{D:.2e}", "D/N:", round(float(D / N), 2))

Section prompt

Copy-only prompts — each action copies a page-grounded prompt to your clipboard. Nothing is sent by this site.

Interactive Demo

Use direct manipulation to connect the explanation to a moving system.

ManipulateChange one control and predict the visible response before reveal.Leave with the observed invariant or a repaired model.

Live Concept Demo

Explore Scaling Laws & Emergent Abilities

The stage is code-native and interactive. Use it to test the explanation against the mechanism.

difficulty 3/5undergraduatecode-aligned

Demo inquiry checkpoint

Manipulate one control and predict the visible change.

01Choose lensTrace a quantity

02ObserveDemo state pending

03GroundName the equation, invariant, or control that explains it.

04CarryNext: RLHF: Reward Modeling + KL-Regularized Policy Optimization

Choose what to inspect in Scaling Laws & Emergent Abilities. This shared fallback is an observation guide, not evidence of learning.

Loading interactive demo...

Use the demo to explore how loss changes as you scale $N$ and $D$ , and why compute-optimal frontiers often prefer more data than you expect.

Section prompt

Copy-only prompts — each action copies a page-grounded prompt to your clipboard. Nothing is sent by this site.

Object

4/4 sections ready

Concept: Scaling Laws & Emergent Abilities

What is the smallest example that makes Scaling Laws & Emergent Abilities click without losing the math?

BeforeScaled Dot-Product Attention & Transformer LayersNow4/4 sections readyTryManipulate one control and predict the visible change.NextRLHF: Reward Modeling + KL-Regularized Policy Optimization

Object contextScaling

StateScaling

2 sources attachedChecking local snapshot

concept:scaling/scaling-laws

Math3 equation objectsL(N, D) \approx L_infinity + a N^{-\alpha} + b D^{-\beta},CodeScaling Laws & Emergent Abilities code witness 1L_inf, a, b = 1.5, 1.0, 1.0 DemoScaling Laws & Emergent Abilities predictionManipulate one control and predict the visible change.

ConceptLearner lens

Scaling Laws & Emergent Abilities

What is the smallest example that makes Scaling Laws & Emergent Abilities click without losing the math?

Mode questionCan I say the mechanism back in one sentence before I reveal anything?

Start with the prediction checkpoint, then compare the reveal to the mental model.

Take this move

Study modes

Keep the object fixed; change the lens.

01PredictCommit before reveal.02ExplainName the invariant.03Ground2 source links nearby.04ExperimentChange one knob.05CarrySave the next question.

Route back through the notebook

Carry the same object through intuition, math, code, and demo.

4/4 sections ready

Carry inScaled Dot-Product Attention & Transformer Layers

Bring the mental model from Scaled Dot-Product Attention & Transformer Layers; this page will reuse it instead of restarting from zero.

Work hereScaling Laws & Emergent Abilities

Empirical power laws that predict how loss and capability improve with parameters, data, and compute, and how to choose compute-optimal training runs.

Carry outRLHF: Reward Modeling + KL-Regularized Policy Optimization

The next edge should feel earned: use the demo prediction here before following RLHF: Reward Modeling + KL-Regularized Policy Optimization.

After The First Pass

Turn the concept into an inspected object.

The lower panels are one second act: keep the object fixed, inspect it visually, check source boundaries, practice transfer, then attach the research question.

ConceptScaling Laws & Emergent AbilitiesScaling

01MechanismInspect motionSee how the idea changes form before adding more explanation.02VisualChoose a cueBind one visible mark to one mathematical role.03SourcesCheck boundary2 source links attached to the object.04ClaimAudit evidenceTreat the page claim as something to verify.05PracticeAttempt transferWrite before feedback so the model has something to repair.06RoomAsk groundedAttach the question to the same exact object.

Mechanism Storyboard

See the idea move before the page explains it

Empirical power laws that predict how loss and capability improve with parameters, data, and compute, and how to choose compute-optimal training runs.

Demo notes open01 / Intuition

Editorial scaling-laws illustration of empirical power-law curves, compute contours, and a highlighted compute-optimal frontier point.

Prediction lens

Start with the picture, metaphor, or geometric mechanism.

Commit first

Before reading further, choose the kind of change Scaling Laws & Emergent Abilities should make visible.

Visual Inquiry

Make the image answer a mathematical question

Empirical power laws that predict how loss and capability improve with parameters, data, and compute, and how to choose compute-optimal training runs.

4/4 stages readyDemo notes connected

Visual cueWhich visible object should carry the first intuition?

Inspection depth2/4

Prediction

Which visible object should carry the first intuition?

Commit first

Pick the cue that should make Scaling Laws & Emergent Abilities easier to reason about before the page gives the answer.

Source Grounding

Canonical references for the mechanism on this page.

Object - ConceptScaling Laws & Emergent AbilitiesQuestion

What is the smallest example that makes Scaling Laws & Emergent Abilities click without losing the math?

concept:scaling/scaling-laws

Boundary

sources: kaplan-2020-scaling-laws, hoffmann-2022-chinchilla

Check

Open the closest source note before trusting the local explanation.

Evidence

2 selected-object sources shown first; 2 references total.

Next move

Audit the claim boundary, then ask from the same selected object.

selected object source · paper · 2020Scaling Laws for Neural Language ModelsKaplan et al.

Located CF editorial boundary

Grounds empirical power-law fits between model size, data, compute, and language-model loss.

Used here as

Kaplan et al. study language-model cross-entropy loss and report power-law scaling with model size, dataset size, and training compute. Hoffmann et al. train over 400 models and find that...

Caveat

This checks empirical pretraining scaling and training-compute allocation, not a proof of intelligence, universal exponents across architectures/datasets, inference-time compute, data-qua...

Open source

selected object source · paper · 2022Training Compute-Optimal Large Language ModelsHoffmann et al.

Located CF editorial boundary

Grounds compute-optimal scaling as a balance between parameter count and training tokens.

Used here as

Kaplan et al. study language-model cross-entropy loss and report power-law scaling with model size, dataset size, and training compute. Hoffmann et al. train over 400 models and find that...

Caveat

This checks empirical pretraining scaling and training-compute allocation, not a proof of intelligence, universal exponents across architectures/datasets, inference-time compute, data-qua...

Open source

Claim Review

Empirical power laws that predict how loss and capability improve with parameters, data, and compute, and how to choose compute-optimal training runs.

Object - ConceptScaling Laws & Emergent AbilitiesQuestion

What is the smallest example that makes Scaling Laws & Emergent Abilities click without losing the math?

concept:scaling/scaling-laws

Boundary

sources: kaplan-2020-scaling-laws, hoffmann-2022-chinchilla

Check

Treat every claim as provisional until source support and a local witness agree.

Evidence

1 structured claim check on this concept.

Next move

Run the prediction or practice transfer before asking for a grounded review.

1 CF editorial source-scope review recorded

Publisher-side editorial review is not independent replication. Claims without it still need exact source-support review. 2 references and 3 local witnesses are available for inspection.

Kaplan et al. fit language-model cross-entropy loss with empirical power laws in model size, dataset size, and training compute; Hoffmann et al. find compute-optimal training scales parameters and tokens together under a fixed budget.

Registered CF editorial boundary

Scaling Laws for Neural Language ModelsGrounds empirical power-law fits between model size, data, compute, and language-model loss.Training Compute-Optimal Large Language ModelsGrounds compute-optimal scaling as a balance between parameter count and training tokens.

Used here as

Kaplan et al. study language-model cross-entropy loss and report power-law scaling with model size, dataset size, and training compute. Hoffmann et al. train over 400 models and find that compute-optimal tra...

Local witness

source-span-kaplan-2020-scaling-laws source-span-hoffmann-2022-chinchilla math-object-1 math-object-2 code-witness-1 interactive-demo

Equation 1

L(N, D) \approx L_\infty + a\,N^{-\alpha} + b\,D^{-\beta},

Equation 2

C \propto N\cdot D.

Caveat

This checks empirical pretraining scaling and training-compute allocation, not a proof of intelligence, universal exponents across architectures/datasets, inference-time compute, data-quality effects, downst...

Review stateCF editorial source-scope reviewClaim metadata: source checkedPublisher-side editorial review only; not independent replication. Check caveats and exact source scope.

Checked arXiv abstracts: Kaplan et al. support cross-entropy loss power laws over model size, dataset size, and training compute; Hoffmann et al. support equal scaling of model size and token count for compute-optimal fixed-budget training. Local math/code/demo instantiate only this scoped pretraining loss and N-D budget view.

Reviewer: codex; reviewed 2026-05-20

Practice · Scaling Laws & Emergent Abilities

Try the idea in your own words

Empirical power laws that predict how loss and capability improve with parameters, data, and compute, and how to choose compute-optimal training runs.

Concept · Current object

Scaling Laws & Emergent Abilities

Source boundary: sources: kaplan-2020-scaling-laws, hoffmann-2022-chinchilla

Object context and links

Scaling

concept:scaling/scaling-laws

Choose a task

Explain the mechanism

For Scaling Laws & Emergent Abilities: What is the smallest example that makes Scaling Laws & Emergent Abilities click without losing the math? Explain your answer, including what changes, why, and which assumption matters.

Your answerNo answer yet

A rough first thought is enough. Your draft stays when you change tasks.

Local to this page session. Not saved after leaving or reloading.

A little help · Explain

Open one hint at a time. These are suggestions, not your answer or a grade.

0 of 3 hints shown for this question.

Clearing your answer does not erase help history. Outside help cannot be verified here.

Where am I stuck? (optional)

Your own description, not an automatic diagnosis

Choose one, or leave this unspecified. Select it again to clear it.

Take your draft to a feedback conversation

No AI feedback runs here. You can copy a prompt to use elsewhere; nothing is sent automatically. Review the text before sharing, and leave out private information.

Write an attempt before copying a feedback prompt.

This draft and any AI response do not establish mastery. Try a different case later without help; this page has not measured that learning.

Grounded object roomClose

Selected object routeAsk from this object; carry one invariant back.sources: kaplan-2020-scaling-laws, hoffmann-2022-chinchilla

ObjectConceptScaling Laws & Emergent Abilities
PredictBefore revealScaling Laws & Emergent Abilities prediction
WitnessCompare codeScaling Laws & Emergent Abilities code witness 1
RoomAsk groundedChecking local snapshot

ConceptScaling Laws & Emergent AbilitiesScaling

Code witness comparisonScaling Laws & Emergent Abilities code witness 1L_inf, a, b = 1.5, 1.0, 1.0 Prediction before revealScaling Laws & Emergent Abilities predictionManipulate one control and predict the visible change.

Grounded room questionWhat is the smallest example that makes Scaling Laws & Emergent Abilities click without losing the math?Checking local snapshot

Research Room

Attach the question to an exact object

Pick the concept, equation, source, code witness, claim, misconception, or demo state before asking for help. The handoff stays grounded to that object.

Next local actionNo local draft saved yet

Open the draft below to save one note and next action in this browser.

conceptScaling

Scaling Laws & Emergent Abilities

Anchored question

What is the smallest example that makes Scaling Laws & Emergent Abilities click without losing the math?

Source boundaryInspect source ids: kaplan-2020-scaling-laws, hoffmann-2022-chinchillaStable content-object key attached

Role lenses for this object

These are fixed, deterministic perspectives derived from the selected object. They do not represent people, community contributions, or independent review.

Learner evidence requestAsk what would make "Scaling Laws & Emergent Abilities" feel predictable rather than familiar.

Assumption

Source ids kaplan-2020-scaling-laws, hoffmann-2022-chinchilla must support the exact object, not just the surrounding topic.

Source-checking summary

Treat this as a mechanism object: connect the definition to one equation, code witness, or demo before broadening the discussion.

Proposed experiment

Ask the learner to perturb one representation, then check whether the same invariant survives in math, code, and demo.

Next action

The learner can state the mechanism in their own words

Evidence4 checks

PredictionChecking carried observation

ActionReady for one action

AILearner handoff ready

Open source object

01PredictionChecking browser-local route memory

02EvidenceChecking for a carried observation

03BoundaryInspect source ids: kaplan-2020-scaling-laws, hoffmann-2022-chinchilla

04Next moveSave one next action

Local action draftNo local draft saved yetExpand only when ready to capture one local next action

Local action draft

This draft stays locally in this browser for concept:scaling/scaling-laws.

Draft noteNext action

No local draft saved.

Evidence to inspect

Source ids to inspect: kaplan-2020-scaling-laws, hoffmann-2022-chinchilla
Definition, prerequisite, and contrast concept links
The equation or code witness that makes the concept operational
One demo state that shows the invariant instead of a slogan

What would resolve this

The learner can state the mechanism in their own words
The learner can name the prerequisite that would repair confusion
The learner can predict how the mechanism changes under one perturbation

Object-attached AI handoff

I am working in Continuous Function's research reading room. Object: concept - Scaling Laws & Emergent Abilities Object key: concept:scaling/scaling-laws Context: Scaling Anchor id: concept/concept-notebook/scaling/scaling-laws Open question: What is the smallest example that makes Scaling Laws & Emergent Abilities click without losing the math? Evidence to inspect: - Source ids to inspect: kaplan-2020-scaling-laws, hoffmann-2022-chinchilla - Definition, prerequisite, and contrast concept links - The equation or code witness that makes the concept operational - One demo state that shows the invariant instead of a slogan Deterministic role lenses for this object: - Boundary: fixed perspectives, not people, community contributions, or independent review - Source-checking summary: Treat this as a mechanism object: connect the definition to one equation, code witness, or demo before broadening the discussion. - Proposed experiment: Ask the learner to perturb one representation, then check whether the same invariant survives in math, code, and demo. - Teach/transfer move: Turn the mechanism into one sentence that predicts a neighboring concept. - Assumptions: - Source ids kaplan-2020-scaling-laws, hoffmann-2022-chinchilla must support the exact object, not just the surrounding topic. - The stable content-object key lets local drafts, prompts, and route memory attach without changing the source page. - The concept explanation is local atlas prose until checked against its math, code, and source support. - Prerequisite gaps should become a repair route, not a reason to leave the object vague. - Role-lens requests: - Learner: ask for "Ask what would make "Scaling Laws & Emergent Abilities" feel predictable rather than familiar." | assumption: Source ids kaplan-2020-scaling-laws, hoffmann-2022-chinchilla must support the exact object, not just the surrounding topic. | next action: The learner can state the mechanism in their own words - Researcher: ask for "Source ids to inspect: kaplan-2020-scaling-laws, hoffmann-2022-chinchilla" | assumption: The stable content-object key lets local drafts, prompts, and route memory attach without changing the source page. | next action: The learner can name the prerequisite that would repair confusion - Experimenter: ask for "Choose one variable or condition to perturb before asking for an explanation." | assumption: The concept explanation is local atlas prose until checked against its math, code, and source support. | next action: The learner can predict how the mechanism changes under one perturbation - Professor: ask for "Find the smallest transferable rule a learner could reuse without the AI." | assumption: Prerequisite gaps should become a repair route, not a reason to leave the object vague. | next action: Teach or transfer: Turn the mechanism into one sentence that predicts a neighboring concept. What would resolve this: - The learner can state the mechanism in their own words - The learner can name the prerequisite that would repair confusion - The learner can predict how the mechanism changes under one perturbation Answer as a careful research tutor: stay source-grounded, separate verified evidence from assumptions, name the relevant math objects, and end with one next action. Current deterministic role lens for this object: - Role lens: Learner - Evidence request: Ask what would make "Scaling Laws & Emergent Abilities" feel predictable rather than familiar. - Assumption to keep visible: Source ids kaplan-2020-scaling-laws, hoffmann-2022-chinchilla must support the exact object, not just the surrounding topic. - Proposed experiment: Ask the learner to perturb one representation, then check whether the same invariant survives in math, code, and demo. - Next action: The learner can state the mechanism in their own words

concept/concept-notebook/scaling/scaling-laws
concept:scaling/scaling-laws

Return to reading

Reading map

In this concept

Manipulate one control and predict the visible change.

Try the demo

Continue with

RLHF: Reward Modeling + KL-Regularized Policy Optimization

Connections and sources

Domain

Scaling

scaling
scaling-laws
chinchilla
compute
emergent-abilities

Prerequisites

Leads To

Object Companion

Ask beside the selected object

Empirical power laws that predict how loss and capability improve with parameters, data, and compute, and how to choose compute-optimal training runs.

Your question

GoalComfortStyleStuck on

Context prompt

You are my AI learning companion for Continuous Function. Current context: Scaling concept. Learning surface: Scaling Laws & Emergent Abilities. What this page says: Empirical power laws that predict how loss and capability improve with parameters, data, and compute, and how to choose compute-optimal training runs. Current section: Intuition, math, code, and interactive demo. Suggested next step: Manipulate one control and predict the visible change.. Learner goal: Understand the idea. Learner comfort level: New to this. Preferred explanation style: Visual first. Task: Explain the central idea in plain language, then restate it with the exact math objects from the page. Answer in a way that helps me learn: ask one clarifying question only if needed, use intuition before notation, and end with one thing I should try on the page.

Copy-only tool: this panel composes a prompt for an AI tool you already use. Nothing you enter here is sent from this page.