Showing 1–18 of 18 results for author: Gilson, R

Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red Teaming

Authors: Mrinank Sharma, Meg Tong, Jesse Mu, Jerry Wei, Jorrit Kruthoff, Scott Goodfriend, Euan Ong, Alwin Peng, Raj Agarwal, Cem Anil, Amanda Askell, Nathan Bailey, Joe Benton, Emma Bluemke, Samuel R. Bowman, Eric Christiansen, Hoagy Cunningham, Andy Dau, Anjali Gopal, Rob Gilson, Logan Graham, Logan Howard, Nimit Kalra, Taesung Lee, Kevin Lin , et al. (18 additional authors not shown)

Abstract: Large language models (LLMs) are vulnerable to universal jailbreaks-prompting strategies that systematically bypass model safeguards and enable users to carry out harmful processes that require many model interactions, like manufacturing illegal substances at scale. To defend against these attacks, we introduce Constitutional Classifiers: safeguards trained on synthetic data, generated by promptin… ▽ More Large language models (LLMs) are vulnerable to universal jailbreaks-prompting strategies that systematically bypass model safeguards and enable users to carry out harmful processes that require many model interactions, like manufacturing illegal substances at scale. To defend against these attacks, we introduce Constitutional Classifiers: safeguards trained on synthetic data, generated by prompting LLMs with natural language rules (i.e., a constitution) specifying permitted and restricted content. In over 3,000 estimated hours of red teaming, no red teamer found a universal jailbreak that could extract information from an early classifier-guarded LLM at a similar level of detail to an unguarded model across most target queries. On automated evaluations, enhanced classifiers demonstrated robust defense against held-out domain-specific jailbreaks. These classifiers also maintain deployment viability, with an absolute 0.38% increase in production-traffic refusals and a 23.7% inference overhead. Our work demonstrates that defending against universal jailbreaks while maintaining practical deployment viability is tractable. △ Less

Submitted 30 January, 2025; originally announced January 2025.

Soliton Solutions of Noncommutative Anti-Self-Dual Yang-Mills Equations

Authors: Claire R. Gilson, Masashi Hamanaka, Shan-Chi Huang, Jonathan J. C. Nimmo

Abstract: We present exact soliton solutions of anti-self-dual Yang-Mills equations for G=GL(N) on noncommutative Euclidean spaces in four-dimension by using the Darboux transformations. Generated solutions are represented by quasideterminants of Wronski matrices in compact forms. We give special one-soliton solutions for G=GL(2) whose energy density can be real-valued. We find that the soliton solutions ar… ▽ More We present exact soliton solutions of anti-self-dual Yang-Mills equations for G=GL(N) on noncommutative Euclidean spaces in four-dimension by using the Darboux transformations. Generated solutions are represented by quasideterminants of Wronski matrices in compact forms. We give special one-soliton solutions for G=GL(2) whose energy density can be real-valued. We find that the soliton solutions are the same as the commutative ones and can be interpreted as one-domain walls in four-dimension. Scattering processes of the multi-soliton solutions are also discussed. △ Less

Submitted 23 July, 2020; v1 submitted 3 April, 2020; originally announced April 2020.

Comments: 20 pages; Jonathan Nimmo sadly passed away on 20 June 2017. We owe section 4 to his unpublished note; v2: minor changes, comments added, references added; version to appear in Special Issue of Journal of Physics A on Integrable Physics and its Connections with Special Functions and Combinatorics

Darboux dressing and undressing for the ultradiscrete KdV equation

Authors: Jonathan J. C. Nimmo, Claire R. Gilson, R. Willox

Abstract: We solve the direct scattering problem for the ultradiscrete Korteweg de Vries (udKdV) equation, over $\mathbb R$ for any potential with compact (finite) support, by explicitly constructing bound state and non-bound state eigenfunctions. We then show how to reconstruct the potential in the scattering problem at any time, using an ultradiscrete analogue of a Darboux transformation. This is achieved… ▽ More We solve the direct scattering problem for the ultradiscrete Korteweg de Vries (udKdV) equation, over $\mathbb R$ for any potential with compact (finite) support, by explicitly constructing bound state and non-bound state eigenfunctions. We then show how to reconstruct the potential in the scattering problem at any time, using an ultradiscrete analogue of a Darboux transformation. This is achieved by obtaining data uniquely characterising the soliton content and the `background' from the initial potential by Darboux transformation. △ Less

Submitted 19 September, 2019; v1 submitted 18 March, 2019; originally announced March 2019.

Comments: 41 pages, 5 figures // Full, unabridged version, including two appendices

Statistics of photon-subtracted and photon-added states

Authors: Stephen M. Barnett, Gergely Ferenczi, Claire R. Gilson, Fiona C. Speirits

Abstract: The subtraction or addition of a prescribed number of photons to a field mode does not, in general, simply shift the probability distribution by the number of subtracted or added photons. Subtraction of a photon from an initial coherent state, for example, leaves the photon statistics unchanged and the same process applied to an initial thermal state increases the mean photon number. We present a… ▽ More The subtraction or addition of a prescribed number of photons to a field mode does not, in general, simply shift the probability distribution by the number of subtracted or added photons. Subtraction of a photon from an initial coherent state, for example, leaves the photon statistics unchanged and the same process applied to an initial thermal state increases the mean photon number. We present a detailed analysis of the effects of the initial photon statistics on those of the state from which the photons have been subtracted or to which they have been added. Our approach is based on two closely-related moment generating functions, one well-established and one that we introduce. △ Less

Submitted 25 June, 2018; originally announced June 2018.

Collision of Multimode Dromions and a Firewall in the Two Component Long Wave Short Wave Resonance Interaction Equation

Authors: R. Radha, C. Senthil Kumar, M. Lakshmanan, C. R. Gilson

Abstract: In this paper, we investigate the two component long wave short wave resonance interaction (2CLSRI) equation and show that it admits the Painleve property. We then suitably exploit the recently developed truncated Painleve approach to generate exponentially localized solutions for the short wave components $S^{(1)}$ and $S^{(2)}$ while the long wave L admits line soliton only. The exponentially… ▽ More In this paper, we investigate the two component long wave short wave resonance interaction (2CLSRI) equation and show that it admits the Painleve property. We then suitably exploit the recently developed truncated Painleve approach to generate exponentially localized solutions for the short wave components $S^{(1)}$ and $S^{(2)}$ while the long wave L admits line soliton only. The exponentially localized solutions driving the short waves $S^{(1)}$ and $S^{(2)}$ in the y direction are endowed with different energies (intensities) and are called "multimode dromions". We also observe that the multimode dromions suffer intramodal inelastic collision while the existence of a firewall across the modes prevents the switching of energy between the modes. △ Less

Submitted 16 April, 2009; originally announced April 2009.

Comments: published in J. Phys. A: Math. Theor. 42, 102002

Dromion solutions of noncommutative Davey-Stewartson equations

Authors: Claire R. Gilson, Susan R. Macfarlane

Abstract: We consider a noncommutative version of the Davey-Stewartson equations and derive two families of quasideterminant solution via Darboux and binary Darboux transformations. These solutions can be verified by direct substitution. We then calculate the dromion solutions of the equations and obtain computer plots in a noncommutative setting. We consider a noncommutative version of the Davey-Stewartson equations and derive two families of quasideterminant solution via Darboux and binary Darboux transformations. These solutions can be verified by direct substitution. We then calculate the dromion solutions of the equations and obtain computer plots in a noncommutative setting. △ Less

Submitted 18 May, 2009; v1 submitted 30 January, 2009; originally announced January 2009.

Comments: 23 pages, 6 figures

Journal ref: J. Phys. A: Math. Theor. 42 (2009) 235202

Bäcklund Transformations and the Atiyah-Ward ansatz for Noncommutative Anti-Self-Dual Yang-Mills Equations

Authors: Claire R. Gilson, Masashi Hamanaka, Jonathan J. C. Nimmo

Abstract: We present Backlund transformations for the noncommutative anti-self-dual Yang-Mills equation where the gauge group is G=GL(2) and use it to generate a series of exact solutions from a simple seed solution. The solutions generated by this approach are represented in terms of quasideterminants. We also explain the origins of all of the ingredients of the Backlund transformations within the framew… ▽ More We present Backlund transformations for the noncommutative anti-self-dual Yang-Mills equation where the gauge group is G=GL(2) and use it to generate a series of exact solutions from a simple seed solution. The solutions generated by this approach are represented in terms of quasideterminants. We also explain the origins of all of the ingredients of the Backlund transformations within the framework of noncommutative twistor theory. In particular we show that the generated solutions belong to a noncommutative version of the Atiyah-Ward ansatz. △ Less

Submitted 16 September, 2009; v1 submitted 8 December, 2008; originally announced December 2008.

Comments: v2: 21 pages, published version

Journal ref: Proc. Roy. Soc. A 465 (2009) 2613-2632

Matrix solutions of a noncommutative KP equation and a noncommutative mKP equation

Authors: C. R. Gilson, J. J. C. Nimmo, C. M. Sooman

Abstract: Matrix solutions of a noncommutative KP and a noncommutative mKP equation which can be expressed as quasideterminants are discussed. In particular, we investigate interaction properties of two-soliton solutions. Matrix solutions of a noncommutative KP and a noncommutative mKP equation which can be expressed as quasideterminants are discussed. In particular, we investigate interaction properties of two-soliton solutions. △ Less

Submitted 8 January, 2009; v1 submitted 10 October, 2008; originally announced October 2008.

Comments: 2 figures

On a direct approach to quasideterminant solutions of a noncommutative modified KP equation

Authors: C. R. Gilson, J. J. C. Nimmo, C. M. Sooman

Abstract: A noncommutative version of the modified KP equation and a family of its solutions expressed as quasideterminants are discussed. The origin of these solutions is explained by means of Darboux transformations and the solutions are verified directly. We also verify directly an explicit connection between quasideterminant solutions of the noncommutative mKP equation and the noncommutative KP equati… ▽ More A noncommutative version of the modified KP equation and a family of its solutions expressed as quasideterminants are discussed. The origin of these solutions is explained by means of Darboux transformations and the solutions are verified directly. We also verify directly an explicit connection between quasideterminant solutions of the noncommutative mKP equation and the noncommutative KP equation arising from the Miura transformation. △ Less

Submitted 22 February, 2008; v1 submitted 23 November, 2007; originally announced November 2007.

Journal ref: J. Phys. A: Math. Theor. 41 (2008) 085202

Bäcklund Transformations for Noncommutative Anti-Self-Dual Yang-Mills Equations

Authors: Claire R. Gilson, Masashi Hamanaka, Jonathan J. C. Nimmo

Abstract: We present Backlund transformations for the noncommutative anti-self-dual Yang-Mills equation where the gauge group is G=GL(2) and use it to generate a series of exact solutions from a simple seed solution. The solutions generated by this approach are represented in terms of quasideterminants and belong to a noncommutative version of the Atiyah-Ward ansatz. In commutative limit, our results coin… ▽ More We present Backlund transformations for the noncommutative anti-self-dual Yang-Mills equation where the gauge group is G=GL(2) and use it to generate a series of exact solutions from a simple seed solution. The solutions generated by this approach are represented in terms of quasideterminants and belong to a noncommutative version of the Atiyah-Ward ansatz. In commutative limit, our results coincide with those by Corrigan, Fairlie, Yates and Goddard. △ Less

Submitted 12 June, 2008; v1 submitted 13 September, 2007; originally announced September 2007.

Comments: 10 pages; version to appear in Glasgow Mathematical Journal

Journal ref: Glasgow Mathematical Journal 51A (2009) 83-93

Quasideterminant solutions of a non-Abelian Hirota-Miwa equation

Authors: C. R. Gilson, J. J. C. Nimmo, Y. Ohta

Abstract: A non-Abelian version of the Hirota-Miwa equation is considered. In an earlier paper [Nimmo (2006) J. Phys. A: Math. Gen. \textbf{39}, 5053-5065] it was shown how solutions expressed as quasideterminants could be constructed for this system by means of Darboux transformations. In this paper we discuss these solutions from a different perspective and show that the solutions are quasi-Plücker coor… ▽ More A non-Abelian version of the Hirota-Miwa equation is considered. In an earlier paper [Nimmo (2006) J. Phys. A: Math. Gen. \textbf{39}, 5053-5065] it was shown how solutions expressed as quasideterminants could be constructed for this system by means of Darboux transformations. In this paper we discuss these solutions from a different perspective and show that the solutions are quasi-Plücker coordinates and that the non-Abelian Hirota-Miwa equation may be written as a quasi-Plücker relation. The special case of the matrix Hirota-Miwa equation is also considered using a more traditional, bilinear approach and the techniques are compared. △ Less

Submitted 9 February, 2007; originally announced February 2007.

On a direct approach to quasideterminant solutions of a noncommutative KP equation

Authors: C. R. Gilson, J. J. C. Nimmo

Abstract: A noncommutative version of the KP equation and two families of its solutions expressed as quasideterminants are discussed. The origin of these solutions is explained by means of Darboux and binary Darboux transformations. Additionally, it is shown that these solutions may also be verified directly. This approach is reminiscent of the wronskian technique used for the Hirota bilinear form of the… ▽ More A noncommutative version of the KP equation and two families of its solutions expressed as quasideterminants are discussed. The origin of these solutions is explained by means of Darboux and binary Darboux transformations. Additionally, it is shown that these solutions may also be verified directly. This approach is reminiscent of the wronskian technique used for the Hirota bilinear form of the regular, commutative KP equation but, in the noncommutative case, no bilinearising transformation is available. △ Less

Submitted 16 January, 2007; v1 submitted 13 January, 2007; originally announced January 2007.

Comments: 11 pages

Maximum Confidence Quantum Measurements

Authors: Sarah Croke, Erika Andersson, Stephen M. Barnett, Claire R. Gilson, John Jeffers

Abstract: We consider the problem of discriminating between states of a specified set with maximum confidence. For a set of linearly independent states unambiguous discrimination is possible if we allow for the possibility of an inconclusive result. For linearly dependent sets an analogous measurement is one which allows us to be as confident as possible that when a given state is identified on the basis… ▽ More We consider the problem of discriminating between states of a specified set with maximum confidence. For a set of linearly independent states unambiguous discrimination is possible if we allow for the possibility of an inconclusive result. For linearly dependent sets an analogous measurement is one which allows us to be as confident as possible that when a given state is identified on the basis of the measurement result, it is indeed the correct state. △ Less

Submitted 5 April, 2006; originally announced April 2006.

Comments: 4 pages, 2 figures

Journal ref: Phys Rev Lett 96, 070401 (2006)

Maximum Fidelity Retransmission of Mirror Symmetric Qubit States

Authors: Kieran Hunter, Erika Andersson, Claire R. Gilson, Stephen M. Barnett

Abstract: In this paper we address the problem of optimal reconstruction of a quantum state from the result of a single measurement when the original quantum state is known to be a member of some specified set. A suitable figure of merit for this process is the fidelity, which is the probability that the state we construct on the basis of the measurement result is found by a subsequent test to match the o… ▽ More In this paper we address the problem of optimal reconstruction of a quantum state from the result of a single measurement when the original quantum state is known to be a member of some specified set. A suitable figure of merit for this process is the fidelity, which is the probability that the state we construct on the basis of the measurement result is found by a subsequent test to match the original state. We consider the maximisation of the fidelity for a set of three mirror symmetric qubit states. In contrast to previous examples, we find that the strategy which minimises the probability of erroneously identifying the state does not generally maximise the fidelity. △ Less

Submitted 25 September, 2002; originally announced September 2002.

Comments: 16 pages, 1 figure, subject of a poster presentation at qcmc02 conference at MIT (MA, USA) in July 2002

Minimum-error discrimination between three mirror-symmetric states

Authors: Erika Andersson, Stephen M. Barnett, Claire R. Gilson, Kieran Hunter

Abstract: We present the optimal measurement strategy for distinguishing between three quantum states exhibiting a mirror symmetry. The three states live in a two-dimensional Hilbert space, and are thus overcomplete. By mirror symmetry we understand that the transformation {|+> -> |+>, |-> -> -|->} leaves the set of states invariant. The obtained measurement strategy minimizes the error probability. An ex… ▽ More We present the optimal measurement strategy for distinguishing between three quantum states exhibiting a mirror symmetry. The three states live in a two-dimensional Hilbert space, and are thus overcomplete. By mirror symmetry we understand that the transformation {|+> -> |+>, |-> -> -|->} leaves the set of states invariant. The obtained measurement strategy minimizes the error probability. An experimental realization for polarized photons, realizable with current technology, is suggested. △ Less

Submitted 17 January, 2002; originally announced January 2002.

Comments: 4 pages, 2 figures

Journal ref: Phys. Rev. A 65, 052308 (2002)

Fidelity and the communication of quantum information

Authors: Stephen M. Barnett, Claire R. Gilson, Masahide Sasaki

Abstract: We compare and contrast the error probability and fidelity as measures of the quality of the receiver's measurement strategy for a quantum communications system. The error probability is a measure of the ability to retrieve {\it classical} information and the fidelity measures the retrieval of {\it quantum} information. We present the optimal measurement strategies for maximising the fidelity gi… ▽ More We compare and contrast the error probability and fidelity as measures of the quality of the receiver's measurement strategy for a quantum communications system. The error probability is a measure of the ability to retrieve {\it classical} information and the fidelity measures the retrieval of {\it quantum} information. We present the optimal measurement strategies for maximising the fidelity given a source that encodes information on the symmetric qubit-states. △ Less

Submitted 5 July, 2001; originally announced July 2001.

Comments: 11 pages, RevTeX, 3 figures(EPSF)

Journal ref: J. Phys. A: Math. Gen. 34(35), 6755 (2001).

Entanglement, Information and Multiparticle Quantum Operations

Authors: Anthony Chefles, Claire R. Gilson, Stephen M. Barnett

Abstract: Collective operations on a network of spatially-separated quantum systems can be carried out using local quantum (LQ) operations, classical communication (CC) and shared entanglement (SE). Such operations can also be used to communicate classical information and establish entanglement between distant parties. We show how these facts lead to measures of the inseparability of quantum operations, a… ▽ More Collective operations on a network of spatially-separated quantum systems can be carried out using local quantum (LQ) operations, classical communication (CC) and shared entanglement (SE). Such operations can also be used to communicate classical information and establish entanglement between distant parties. We show how these facts lead to measures of the inseparability of quantum operations, and argue that a maximally-inseparable operation on 2 qubits is the SWAP operation. The generalisation of our argument to N qubit operations leads to the conclusion that permutation operations are maximally-inseparable. For even N, we find the minimum SE and CC resources which are sufficient to perform an arbitrary collective operation. These minimum resources are 2(N-1) ebits and 4(N-1) bits, and these limits can be attained using a simple teleportation-based protocol. We also obtain lower bounds on the minimum resources for the odd case. For all $N{\geq}4$, we show that the SE/CC resources required to perform an arbitrary operation are strictly greater than those that any operation can establish/communicate. △ Less

Submitted 28 June, 2000; v1 submitted 23 June, 2000; originally announced June 2000.

Entanglement and Collective Quantum Operations

Authors: Anthony Chefles, Claire R. Gilson, Stephen M. Barnett

Abstract: We show how shared entanglement, together with classical communication and local quantum operations, can be used to perform an arbitrary collective quantum operation upon N spatially-separated qubits. A simple teleportation-based protocol for achieving this, which requires 2(N-1) ebits of shared, bipartite entanglement and 4(N-1) classical bits, is proposed. In terms of the total required entang… ▽ More We show how shared entanglement, together with classical communication and local quantum operations, can be used to perform an arbitrary collective quantum operation upon N spatially-separated qubits. A simple teleportation-based protocol for achieving this, which requires 2(N-1) ebits of shared, bipartite entanglement and 4(N-1) classical bits, is proposed. In terms of the total required entanglement, this protocol is shown to be optimal for even N in both the asymptotic limit and for `one-shot' applications. △ Less

Submitted 16 June, 2000; v1 submitted 16 March, 2000; originally announced March 2000.