GHOSTS: graduate-level natural-language mathematics prompts from the paper 'Mathematical Capabilities of ChatGPT' (Frieder et al., 2023). Each prompt is posed to a language model and the model's answer is graded 1-5 by a human mathematician (higher = better). Subjects are the two ChatGPT snapshots (2023-01-09, 2023-01-30) on the full GHOSTS set and GPT-4 on the miniGHOSTS subset; items span definition retrieval, MATH-dataset problems, textbook exercises (Munkres, Rudin, Durrett), olympiad problems, proof completion and symbolic integration.
Fit to width. Hover for subject & item; click a cell for details.
lowhighUnobserved
Scale:{1, 2, 3, 4, 5}
Sample items
Item 1·100% solve rate
For 0<x<∞,
Γ(x)=∫0∞tx−1e−tdt.
The integral converges for these x. One can show the functional equation
Γ(x+1)=xΓ(x).
What is the name of the result used to prove this equation?
How subjects answered
ChatGPT (2023-01-09)score 1
The functional equation
Γ(x+1)=xΓ(x)
is known as the "recursion formula" or "recursion relation" for the Gamma function.
ChatGPT (2023-01-30)score 1
The functional equation Γ(x+1)=xΓ(x) is known as the "Gamma Function Recurrence Relation."
Item 2·200% solve rate
How do you call a function such that its Fourier transform is compactly supported?
How subjects answered
ChatGPT (2023-01-09)score 2
A function whose Fourier transform is compactly supported is called a 'Schwartz function.' The space of all Schwartz functions is also known as the 'Schwartz space.'
ChatGPT (2023-01-30)score 2
A function whose Fourier transform is compactly supported is often referred to as a "Schwartz function". The term "Schwartz function" refers to functions that belong to the Schwartz space, which is a space of rapidly decreasing smooth functions that are well-behaved in both the time and frequency domains. Schwartz functions have numerous useful properties and appear frequently in mathematics and signal processing.
Item 3·200% solve rate
[GHOSTS math problem — prompt text withheld in dataset] category: Olympiad Problem Solving; source: Arthur_Engel Problem_Solving_Strategies Springer_Verlag New-York 1998 p.27 ex16; MSC: 00A07
How subjects answered
ChatGPT (2023-01-09)score 2
It does not matter on which square the -1 is initially placed because any subsquare of size a>1 can be chosen to reverse the signs of all the squares within it. This means that the final configuration of all +1s can be reached regardless of the initial placement of the -1.
ChatGPT (2023-01-30)score 2
It is not possible to reach +1 on each square of a 5×5 chessboard by reversing the signs of a×a subsquares with a>1. No matter which square has the value -1, it will always be impossible to change it to +1 by reversing the signs of larger subsquares. This is because the total number of squares on the chessboard is odd, and so it is not possible to change an odd number of -1s to +1s.
Item 4·200% solve rate
Find the radius of the circle with equation x2+8x+y2−6y=0.
How subjects answered
ChatGPT (2023-01-09)score 2
To find the radius of the circle, we can complete the square to put the equation in standard form: (x+4)2+(y−3)2=r2.
We can see that the center of the circle is (−4,3), and the radius is r.
Therefore the radius of this circle is r.
ChatGPT (2023-01-30)score 2
The equation x2+8x+y2−6y=0 is not the equation of a circle in standard form, which is of the form (x−h)2+(y−k)2=r2. To convert this equation to standard form, we need to complete the square in both x and y.
Completing the square in x:
x2+8x=x2+8x+64=(x+4)2−64
Completing the square in y:
y2−6y=y2−6y+9=(y−3)2−9
Substituting the above results into the original equation, we get:
(x+4)2−64+(y−3)2−9=0
(x+4)2+(y−3)2=73
This is the equation of a circle with center (−4,3) and radius 73. Thus, the radius of the circle is 73.
Item 5·250% solve rate
Find all values of x,0≤x≤2π, which satisfy
[\sin^2 x + \cos x + 1 = 0.]Enter all the solutions, separated by commas.
How subjects answered
ChatGPT (2023-01-09)score 3
The given equation can be rewritten as sin2x+cosx+1=(sinx+1)(sinx+cosx)=0. Thus, the solutions are x=arcsin(−1)+2npi and x=arccos(−1)+2npi, where n is an integer. The values of x between 0 and 2pi are:
So the solutions are 23pi+2npi and 2pi+2npi where n is an integer.
ChatGPT (2023-01-30)score 2
We know that sin2x+cos2x=1, so
[\sin^2 x + \cos x + 1 = 1 + \cos x + 1 = 2.]Thus, no such value of x exists, so the only solution is .
Item 6·300% solve rate
For what real value of k is 413−131 a root of 2x2−13x+k?
How subjects answered
ChatGPT (2023-01-09)score 3
A root of the quadratic equation 2x2−13x+k is a value of x that makes the equation equal to zero. So, we can set the equation equal to zero and substitute the given value for x to find k.
2(413−131)2−13(413−131)+k=0
Solving for k, we get:
k=16169−26131+131=16300−26131
Therefore, the real value of k is 16300−26131
ChatGPT (2023-01-30)score 3
If 413−131 is a root of 2x2−13x+k, then by the Factor Theorem,
2(413−131)2−13(413−131)+k=0.
Expanding the left side and simplifying gives
16131−26131+169−413⋅13−13131+k=0,
which simplifies further to
16304−26131+k=0.
Therefore,
k=−16304−26131=1626131−304.
Item 7·300% solve rate
What is the shortest distance between the circles defined by x2−24x+y2−32y+384=0 and x2+24x+y2+32y+384=0?
How subjects answered
ChatGPT (2023-01-09)score 3
The shortest distance between the two circles is the distance between the centers of the circles minus the sum of their radii. The two circles defined by the equations x2−24x+y2−32y+384=0 and x2+24x+y2+32y+384=0 are centered at (12,16) and (−12,−16), respectively. The radius of each circle can be found by completing the square and setting the equation to zero, which gives a radius of 16. Therefore, the distance between the centers of the circles is 122+(−16−16)2=144+256=400=20. Subtracting the sum of the radii, 16+16 = 32, from the distance between the centers gives 20-32 = -12. Since distance cannot be negative, the shortest distance between …
ChatGPT (2023-01-30)score 3
The shortest distance between two circles can be found by finding the perpendicular bisector of the line segment connecting the closest points on the two circles.
One way to find the distance is to first find the center and radius of each circle, then find the line that bisects the line segment connecting the centers and use the Pythagorean Theorem to find the distance.
Let the first circle be C1 with center (x1,y1) and radius r1, and the second circle be C2 with center (x2,y2) and radius r2. The distance d between the centers of the two circles is given by:
d = √((x2-x1)^2 + (y2-y1)^2)
The equation of the perpendicular bisector of the line segment connecting the two centers is given …
Item 8·350% solve rate
We want to show that any open subset of R is a countable union of disjoint open intervals. Let U be a non-empty open subset of R. For x,y∈U define x∼y iff [min{x,y},max{x,y}]⊆U. It’s easily checked that ∼ is an equivalence relation on U whose equivalence classes are pairwise disjoint open intervals in R. (The term interval here includes unbounded intervals, i.e., rays.) Let I be the set of ∼-classes. Can you finish the proof?
How subjects answered
ChatGPT (2023-01-30)score 4
Yes. Since U is non-empty and open, for each x∈U, there exists an open interval (a,b) such that x∈(a,b)⊆U. It follows that the collection I of pairwise disjoint open intervals covers U and is therefore a countable union of disjoint open intervals. Thus, any open subset of R is a countable union of disjoint open intervals.
ChatGPT (2023-01-09)score 3
Since ∼ is an equivalence relation on U, the equivalence classes form a partition of U. Since U is open, each equivalence class is also open, and the classes are pairwise disjoint. Therefore, U can be expressed as a countable union of disjoint open intervals, where each interval corresponds to a single equivalence class in I. This completes the proof that any open subset of R is a countable union of disjoint open intervals.
Item 9·400% solve rate
Let u and v be vectors such that ∥u∥=∥v∥=2 and u⋅v=−1. If θ is the angle between the vectors u+v and 2u−v, then find cosθ.
Solution: Note that
\begin{align*}
(\mathbf{u} + \mathbf{v}) \cdot (2 \mathbf{u} - \mathbf{v}) &= 2 \mathbf{u} \cdot \mathbf{u} + \mathbf{u} \cdot \mathbf{v} - \mathbf{v} \cdot \mathbf{v} \
&= 2 \cdot 2^2 + (-1) - 2^2 \
&= 3.
\end{align*}Also,
\begin{align*}
|\mathbf{u} + \mathbf{v}| &= \sqrt{(\mathbf{u} + \mathbf{v}) \cdot (\mathbf{u} + \mathbf{v})} \
&= \sqrt{\mathbf{u} \cdot \mathbf{u} + 2 \mathbf{u} \cdot \mathbf{v} + \mathbf{v} \cdot \mathbf{v}} \
&= \sqrt{2^2 + 2(-1) + 2^2} \
&= \sqrt{6},
\end{align*} <MISSING>
Fill in the missing steps.
For v=0−41 and w=22−1, compute projwv.
Solution: We have that
\begin{align*}
\text{proj}_{\bold{w}} \bold{v} &= \frac{\bold{v} \cdot \bold{w}}{\bold{w} \cdot \bold{w}} \bold{w} \
&= <MISSING>
Fill in the missing steps.
How subjects answered
GPT-4score 5
To find the projection of vector v onto vector w, we can follow these steps:
Compute the dot product of v and w: v⋅w
Compute the dot product of w with itself: w⋅w
Divide the dot product of v and w by the dot product of w and itself: w⋅wv⋅w
Multiply the result from step 3 by the vector w.
Now let's compute the missing steps:
To fill in the missing steps, we first need to compute the dot product of v and w, which is (0)(2)+(−4)(2)+(1)(−1)=−6. Next, we need to compute the dot product of w and itself, which is (2)(2)+(2)(2)+(−1)(−1)=9. Finally, we can substitute these values into the formula for the projection: \text{proj}_{\bold{w}} \bold{v} = \frac{-6}{9} \bold{w} = \begin{pmatrix} -4/3 \ -4/3 \ 2/3 end{pmatrix}
How would you call a topological space, where the topology is generated by a family of seminorms?
How subjects answered
ChatGPT (2023-01-09)score 5
A topological space where the topology is generated by a family of seminorms is called a locally convex topological vector space.
ChatGPT (2023-01-30)score 5
A topological space generated by a family of seminorms is usually referred to as a locally convex topological vector space.
GPT-4score 5
A topological space whose topology is generated by a family of seminorms is typically referred to as a seminormed space or seminormed vector space.
However, this is slightly more intricate than it seems. A single seminorm doesn't usually generate a topology that's worth discussing on its own. If you have a vector space and a single seminorm, that's usually not enough to define a topological vector space because a seminorm is allowed to be zero at more places than just the origin.
On the other hand, a family of seminorms can generate a topology on the vector space. This family of seminorms induces a topology known as the initial topology or the weakest topology which makes all the …
Item 12·500% solve rate
Please give a formal definition of a continuous function
How subjects answered
ChatGPT (2023-01-09)score 5
A continuous function is a function for which, given any two input values, the output values can be made arbitrarily close by restricting the input values to be sufficiently close to each other. In other words, for any two input values x and y, and for any positive number ε, there exists a positive number δ such that if the distance between x and y is less than δ, then the distance between f(x) and f(y) is less than ε. This definition uses the concept of a metric space, which defines a notion of distance between elements of a set.
ChatGPT (2023-01-30)score 5
A continuous function is a function in which for any two points in the function's domain, the function maps nearby points in the domain to nearby points in the range, preserving the relationship between input and output. In mathematical terms, a function f from a set X to a set Y is said to be continuous at a point x in X if for every ε > 0, there exists a corresponding δ > 0 such that for every x' in X, if |x' - x| < δ, then |f(x') - f(x)| < ε.